Judge SACK dissents in part in a separate opinion.
Defendants-Appellants Sanjay Kumar and Stephen Richards appeal from separate judgments of conviction by the district court (I. Leo Glasser, Judge), pursuant to their guilty pleas to several counts of conspiracy, securities and wire fraud, obstruction of justice, and perjury. After accepting their pleas, the district court calculated defendants’ Guidelines ranges for them fraud and obstruction offenses pursuant to the Sentencing Guidelines Manual (“Guidelines”) in effect at the time of their sentencings, and sentenced Kumar and Richards to non-Guidelines sentences of imprisonment of 144 months and 84 months, respectively, and ordered restitution payments of $800 million and $29 million, respectively.
On appeal, Richards challenges his conviction for obstruction of justice, arguing that the indictment failed to properly charge him with that offense. In addition, Richards attacks his guilty plea to all counts as constitutionally infirm because it resulted from undue coercion by the government. Both defendants argue that, by calculating their Guidelines range according to the Guidelines in effect at sentencing, instead of at the time the fraud offenses were committed, the district court sentenced them in violation of the Ex Post Facto clause. They also claim that the district court improperly denied them ac*617eeptance of responsibility credit, and that its orders of restitution were based on erroneous loss calculations.
We find no infirmity in Richards’s conviction, in the district court’s application of the 2005 version of the Guidelines,1 in the district court’s loss determination, or in the denial of Kumar’s request for acceptance of responsibility credit at sentencing. We conclude, however, that the district court erroneously denied Richards a reduction on the basis of his acceptance of responsibility and remand for resentencing on that basis.
BACKGROUND
The following facts are material to this appeal.
Kumar joined Computer Associates (“CA”), a publicly traded corporation, in August 1987 and was elevated to CEO in August of 2000 and to Chairman of the Board of Directors in 2002. Richards joined CA in 1988 and became Head of North American Sales in 1999. During the tenure of both defendants, CA engaged in a fraudulent accounting practice known as the “35-day month,” whereby CA backdated contracts executed in the first few days of a financial quarter to recognize that revenue in the prior quarter. The purpose of the 35-day month practice, which began in the 1980s under Kumar’s predecessor, was to deceive investors into believing that the company had met or exceeded its quarterly earnings estimates.
In February 2002, the United States Attorney’s Office (“USAO”) and Securities and Exchange Commission (“SEC”) began a joint investigation into the 35-day month practice as contravening both accounting principles and federal securities law. As part of its investigation, government investigators sought witness statements from CA personnel. On June 9, 2003, the USAO and SEC requested that CA conduct its own internal investigation into the practice and give the USAO and SEC “direct access” to company employees. CA’s outside counsel advised CA to comply fully with the investigation.
On August 25, 2003, the SEC subpoenaed the testimony of ten individuals associated with CA, including Kumar and Richards. The SEC interviews were held at the USAO office in Central Islip, New York. On September 22, 2003, prosecutors and SEC staff told Richards’s counsel that Richards was a target of their criminal investigation, and the SEC reiterated its demand that Richards comply with the subpoena for his testimony. In early October 2003, CA told Richards that he would be terminated if he didn’t comply with the subpoena. On October 22, 2003, CA’s outside counsel interviewed Richards, and the following day, Richards testified before the SEC. Richards falsely denied knowledge of the 35-day month practice during both meetings and in his testimony.
On September 22, 2004, CA entered into a deferred prosecution agreement with the USAO and a civil settlement with the SEC. The following day, an indictment charging Richards and Kumar was unsealed. A superceding indictment was filed on May 17, 2005. By this indictment, Richards was charged with both conspiracy to commit, and substantive counts of, securities and wire fraud, as well as filing false public statements with the SEC and perjury. Richards was also charged under 18 U.S.C. § 1512(c) with obstruction of justice arising out of his false exculpatory statements to CA’s counsel and the SEC.
The superceding indictment charged Kumar with both conspiracy to commit, and substantive counts of, securities and wire fraud, as well as filing false public state*618ments with the SEC and making false statements to the FBI. Kumar was also charged with obstruction of justice. However, Kumar’s obstruction charge (also brought pursuant to 18 U.S.C. § 1512(c)) arose out of different conduct from that of Richards. Specifically, the government alleged that Kumar, in an effort to cover up the existence of the 35-day month practice, lied to CA’s outside counsel, instructed CA’s general counsel to coach CA employees to lie, authorized CA’s general counsel to pay a $3.7 million bribe to an individual to procure his silence, and lied to FBI agents and others during his interview at the USAO’s office.
Both Richards and Kumar made various motions to dismiss the charges against them. Relevant to this appeal, first they unsuccessfully moved to dismiss the obstruction charges, arguing that their oral statements to government investigators were beyond the reach of 18 U.S.C. § 1512(c), which they claimed was confined to documentary evidence.
Next, Richards moved to suppress his false statements to CA’s outside counsel and the SEC, claiming that they were coerced in violation of the Fifth Amendment. Without addressing the merits, the district court denied the motion on the basis that Richards had not shown “good cause” for failing to timely file the motion to suppress pursuant to the court-ordered deadline.
In April 2006, both defendants pled guilty to all charges and the Probation Department prepared a Presentence Report (“PSR”) for each defendant. Both PSRs calculated defendants’ Guidelines ranges based on the instructions provided in the 2005 Sentencing Guidelines, notwithstanding the fact that the 35-day month practice—the basis for the securities fraud charges—ended in 2000. Richards’s PSR arrived at an offense level of 50 and a Guidelines range of life imprisonment, and Kumar’s PSR recommended an offense level of 51 and a Guidelines range of life imprisonment. Both PSRs calculated losses to the public resulting from the 35-day month practice to exceed $400 million.
In August 2006, the defendants submitted objections to the Guidelines calculations in the PSR. The defendants argued that application of the 2005 Sentencing Guidelines (effective November 2005) to the securities fraud offenses instead of the 1998 Sentencing Guidelines (effective November 1998) would violate the Ex Post Facto clause because the 2005 Guidelines contained several significant enhancements for securities fraud that were not in effect when the fraud offenses were committed.2 The defendants further objected to the $400 million victim loss calculation as overinflated.3
In November 2006, the district court sentenced both defendants under the 2005 *619Guidelines. At sentencing, the district court rejected both defendants’ objections to their PSRs, finding that (1) because the' Guidelines were advisory, the Ex Post Facto clause was not implicated by sentencing decisions, and (2) the $400 million figure was not erroneous, as it was based on the persuasive testimony of the government’s expert witness. The district court also rejected an application by both defendants for a two-level reduction to the Guidelines’ base offense level for acceptance of responsibility. The district court found that Kumar had not sufficiently accepted responsibility for his crimes to warrant the departure and that Richards’s acceptance of responsibility was “[unjtimely” because he pled guilty “just ... two weeks before the trial was to begin.” See Richards Sentencing Tr. 15:14, 16:1-2, Nov. 14, 2006; see also Richards Sentencing Tr. 16:5-6 (noting that, in light of Richards’s non-Guidelines sentence, rejection of the two-level reduction “ma[de] very little difference insofar as what [his] sentence [was] going to be”).4
The district court then decided to impose non-Guidelines sentences for both defendants, sentencing downwardly from the Guidelines’ calculation of life imprisonment. The court found that “[t]o impose th[e] sentenced] [recommended by the Guidelines] in this case would shock the conscience of this Court ... [and] the conscience of the reasonable person.” Kumar Sentencing Tr. 65:24-66:1, Nov. 2, 2006; see also Richards Sentencing Tr. 25:5-9. Accordingly, the court sentenced Kumar and Richards to imprisonment terms of twelve years and seven years, respectively.
Defendants appealed.
DISCUSSION
I. Was Richards Properly Convicted Of Obstruction Of Justice?
Richards claims that he was “erroneously convicted” of obstruction of justice because 18 U.S.C. § 1512(c),5 the statute under which he was charged, only covers the destruction of physical evidence, and not the charged conduct of “providing false oral responses under oral questioning from CA’s outside counsel and the SEC.”6 Richards Br. 18. The government argues that Richards’s challenge is barred by his guilty plea, and that, in any event, § 1512(c)(2) covers all obstructive conduct, not just destruction of physical evidence. While the government may be correct that § 1512(c)(2) applies to testimonial evidence, we need not reach this issue of first impression in this circuit, because we find *620that any defect asserted is non-jurisdictional and was waived by the guilty plea.
A plea of guilty “waive[s] any and all non-jurisdictional defects in the indictment.” United States v. Moloney, 287 F.3d 236, 239 (2d Cir.2002). To challenge the court’s jurisdiction, “the defendant who has pleaded guilty must establish that the face of the indictment discloses that the count or counts to which he pleaded guilty failed to charge a federal offense.” Hayle v. United States, 815 F.2d 879, 881 (2d Cir.1987). Thus, to attack a conviction post-plea, a defendant must establish that the district court lacked the “power to entertain the prosecution.” Id. at 882; see United States v. Cotton, 535 U.S. 625, 630, 122 S.Ct. 1781, 152 L.Ed.2d 860 (2002) (defining “jurisdiction” as “the courts’ statutory or constitutional power to adjudicate the case”).
Richards claims that he has met this burden by showing that § 1512(c)(2) does not proscribe false testimony, thereby establishing that he did not violate a federal statute. However, regardless of whether Richards’s false testimony violated § 1512(c)(2), it plainly violated § 1503(a), and the indictment charges at least that offense. Because the indictment created federal jurisdiction under § 1503(a), Richards’s claim has no jurisdictional significance and thus is waived by his guilty plea.
Section 1503(a) provides, in relevant part, that “[w]hoever corruptly ... influences, obstructs, or impedes, or endeavors to influence, obstruct, or impede, the due administration of justice” is guilty of a criminal violation. 18 U.S.C. § 1503(a). In United States v. Aguilar, 515 U.S. 593, 598, 115 S.Ct. 2357, 132 L.Ed.2d 520 (1995), the Supreme Court interpreted this subsection as a “catchall” intended to prohibit all obstructive behavior. See id. (referring to subdivision four of § 1503(a) as “far more general in scope” than the remainder of the subsection). Courts in this circuit have likewise given § 1503’s omnibus clause a generally non-restrictive reading. See, e.g., United States v. Rosner, 352 F.Supp. 915, 919 (S.D.N.Y.1972) (noting that the omnibus clause “embraces the widest variety of conduct that impedes the judicial process”); see also United States v. Solow, 138 F.Supp. 812, 814 (S.D.N.Y.1956) (characterizing the omnibus provision as “all-embracing”). Other circuits’ readings of § 1503’s omnibus clause have been as broad as, or even broader than, the reading in this circuit.7
However, § 1503(a) only prohibits false testimony that has a direct “nexus” to an official government proceeding. Aguilar, 515 U.S. at 600, 115 S.Ct. 2357. In endorsing a nexus test, the Supreme Court stated that
[t]he action taken by the accused must be with an intent to influence judicial or grand jury proceedings; it is not enough that there be an intent to influence some ancillary proceeding, such as an investigation independent of the court’s or grand jury’s authority.... [T]he act must have a relationship in time, causation, or logic with the judicial proceedings. In other words, the endeavor must have the natural and probable effect of interfering with the due administration of justice.... [I]f the defendant lacks knowledge that his actions are likely to affect the judicial proceeding, he lacks the requisite intent to obstruct.
Id. at 599, 115 S.Ct. 2357 (internal citations and quotation marks omitted). The nexus *621limitation is “best understood as an articulation of the proof of wrongful intent that will satisfy the mens rea requirement of ‘corruptly’ obstructing or endeavoring to obstruct.” United States v. Quattrone, 441 F.3d 153, 170 (2d Cir.2006). Thus, statements made to investigating agents who might or might not testify before a grand jury are insufficient to violate § 1503(a), because false testimony given in anticipation of a purely hypothetical judicial proceeding is not covered by § 1503(a). See United States v. Bruno, 383 F.3d 65, 87-88 (2d Cir.2004); see also United States v. Schwarz, 283 F.3d 76, 109 (2d Cir.2002). Nevertheless, a defendant does not need to know with certainty that his conduct would affect judicial proceedings, nor does his conduct need to actually obstruct justice. Instead, the defendant’s conduct must only have the “natural and probable effect of interfering with the due administration of justice.” Aguilar, 515 U.S. at 599, 115 S.Ct. 2357 (internal quotation marks omitted). Section 1503(a) applies where a defendant “intenfds] to obstruct justice” but “is foiled in some way.” Id. at 601-02, 115 S.Ct. 2357.
Here, Richards’s conduct easily falls within the ambit of § 1503(a). Richards lied to the SEC in an attempt to “impede” the SEC’s ongoing investigation, which Richards knew was not simply an agency fishing expedition or an “ancillary proceeding” that might or might not result in criminal proceedings. Id. at 599, 115 S.Ct. 2357. Richards concedes that he knew that criminal proceedings were not merely possible when he made his false statements, but that the government “predicted” that criminal charges would be brought against him prior to his SEC interview. See Richards Br. 10; see also Post-Argument Letter from Stephen Richards to the Court, dated Sept. 26, 2008 (“Richards Letter”), at 1 (“Richards testified ... at the [USAO] under the ... belief that admitting knowledge of the 35 Day Month would incriminate him in securities fraud.”). Richards thus plainly “entertained ... expectations” during his interview at the USAO that his false statements to the SEC would impede impending criminal proceedings brought against him. Schwarz, 283 F.3d at 109; see United States v. Cueto, 151 F.3d 620, 634 (7th Cir.1998) (finding nexus to exist where defendant’s conduct, which “predate[d] the empaneling of the grand jury, ... corruptly endeavored to obstruct the due administration of justice”).8 “[I]t is *622clear that if the representations” made by-Richards “had been believed, the grand jury would have been thrown completely off the trail that it was pursuing with respect to” Richards. United States v. Jespersen, 65 F.3d 993, 1001 (2d Cir.1995). Therefore, there was a sufficiently close “relationship in time, causation, [and] logic” between Richards’s conduct and likely judicial proceedings to satisfy § 1503(a)’s nexus requirement. United States v. Reich, 479 F.3d 179, 186 (2d Cir.2007) (internal quotation marks omitted).
In addition, the indictment fully apprised Richards of the elements of his offense under § 1503(a). Specifically, the indictment stated that
Richards well knew and believed that certain of the statements he made during the interviews were false and that he otherwise concealed during the interviews information which he knew to be material to the Government Investigations. Richards further well knew, and in fact intended, that his false statements and concealment of material information would have the effect of obstructing and impeding the Government Investigations.
As the government correctly notes, “[t]he citation of a statutory section number ... is not a part of the offense, and ... an allegedly erroneous statutory citation is not a jurisdictional defect.” Gov’t Br. 11; see Fed.R.Crim.P. 7(c)(3) (“Unless the defendant was misled and thereby prejudiced, neither an error in a citation nor a citation’s omission is a ground ... to reverse a conviction.”). The failure of the indictment to specify § 1503(a), as opposed to § 1512(c)(2), is therefore not grounds for vacating Richards’s guilty plea.9
Richards’s remaining arguments, the government also correctly notes, are “mere corollaries of [his] ... claim that [he] pleaded to a ‘non-offense.’ ” Gov’t Br. 13-14. For example, Richards argues that his guilty plea should be vacated because he was “misinformed of the elements of [his] crime,” since, according to Richards, § 1512(c)(2) does not reach testimonial evidence. Richards Br. 19 (citing Bousley v. United States, 523 U.S. 614, 618, 118 S.Ct. 1604, 140 L.Ed.2d 828 (1998)). But Richards’s argument rests on the flawed premise that the indictment charged him with a “nonoffense.” That is not the case here; at most, the charge should have been brought under § 1503(a) instead of § 1512(c)(2), and the indictment fully stat*623ed the elements of an offense under that section. Vacating Richards’s plea is therefore unnecessary, as Richards does not “stand[ ] convicted of an act that the law does not make criminal.” Bousley, 523 U.S. at 620, 118 S.Ct. 1604 (internal quotation marks omitted). In any event, because there was no jurisdictional defect in Richards’s obstruction charge, his guilty plea waived any ancillary claims based on that charge.
II. Does Richards’s Guilty Plea Bar His Coercion Claim?
Next, Richards argues that the government violated the Fifth Amendment’s prohibition on government-compelled testimony by placing improper pressure on CA to cooperate in the government’s investigation that resulted in CA’s insistence that Richards either testify before the SEC or be terminated. The district court rejected Richards’s motion to suppress his testimony, which was admittedly false and became the basis for the obstruction of justice charge to which Richards pled guilty, as untimely in light of the district court’s schedule for motions. On appeal, Richards argues that his guilty plea should not bar his coercion claim because review of his claim is necessary “to preserve the integrity of the judicial process.” Richards Br. 32. Richards’s coercion claim is easily resolved because the Fifth Amendment does not protect false testimony.
As previously noted, a plea of guilty “waive[s] any and all non-jurisdictional defects in the indictment.” Moloney, 287 F.3d at 239. However, this court may overturn a guilty plea on involuntariness grounds when the defendant shows that his plea “was substantially motivated by a coerced confession.” United States ex rel. Ross v. McMann, 409 F.2d 1016, 1023 (2d Cir.1969) (en banc), vacated on other‘ grounds sub nom. McMann v. Richardson, 397 U.S. 759, 90 S.Ct. 1441, 25 L.Ed.2d 763 (1970). In addition, a guilty plea that is otherwise “voluntary and intelligent” may be overturned if it contains “constitutional violations” that are “logically inconsistent with the valid establishment of factual guilt.” Menna v. New York, 423 U.S. 61, 62 n. 2, 96 S.Ct. 241, 46 L.Ed.2d 195 (1975) (per curiam).
Richards’s coercion claim is not viable, because it is based on a fatally flawed premise: that false statements—whether or not made under coercive circumstances—are protected by the Fifth Amendment. To the contrary, the Supreme Court has repeatedly held that the Fifth Amendment does not “eonfer[] a privilege to lie.” Brogan v. United States, 522 U.S. 398, 404, 118 S.Ct. 805, 139 L.Ed.2d 830 (1998). “[Pjroper invocation of the Fifth Amendment privilege against compulsory self-incrimination allows a witness to remain silent, but not to swear falsely....” United States v. Apfelbaum, 445 U.S. 115, 117, 100 S.Ct. 948, 63 L.Ed.2d 250 (1980); see also United States v. Wong, 431 U.S. 174, 180, 97 S.Ct. 1823, 52 L.Ed.2d 231 (1977); Bryson v. United States, 396 U.S. 64, 72, 90 S.Ct. 355, 24 L.Ed.2d 264 (1969). Richards’s coercion claim is based entirely on perjured testimony, and thus, is unsupportable. See, e.g., Scher v. Nat’l Ass’n of Sec. Dealers, Inc., 386 F.Supp.2d 402, 409 (S.D.N.Y.2005) (finding plaintiff could not state a claim for “supposed deprivation of her constitutional rights” where she had perjured herself before the National Association of Securities Dealers); see also United States v. Nanni, 59 F.3d 1425, 1431 (2d Cir.1995) (negative inferences may not be drawn from immunized exculpatory testimony “except to the extent that the testimony amounts to perjury”).
Richards is attempting to turn the “dilemma” faced by individuals who are im*624properly coerced into incriminating themselves or who risk a negative inference from their silence, into a “trilemma” whereby an individual feels he must either testify truthfully, risk the inference, or lie under oath. See Brogan, 522 U.S. at 404, 118 S.Ct. 805. The Supreme Court rejected this “trilemma,” which is “wholly of the guilty suspect’s own making,” as insufficient to implicate the Fifth Amendment’s protection. Id. The government may have been provided “the leverage to secure [Richards’s] guilty plea” by Richards’s lies, see Richards Reply Br. 25, but Richards’s attempt to draw a sort of constitutional equivalency between a “coerced confession” and a coerced lie, see Richards Reply Br. 13 (quoting Higazy v. Templeton, 505 F.3d 161, 171 (2d Cir.2007)), is exactly the type of “compassion inflation” that the Supreme Court has declared outside the purview of the Fifth Amendment, Brogan, 522 U.S. at 404, 118 S.Ct. 805. Thus, it was not “ironic” that the government used Richards’s false statements in its obstruction charge against him, see Richards Reply Br. 25, but simply a predictable consequence of his mendacity.
Richards argues that the rule denying constitutional protection to coerced false testimony does not apply to this case because the government acted “covert[ly]” in securing his testimony, by “compell[ing][him] to speak without disclosing ... its role in his dilemma to speak or be fired.” Richards Letter at 1. But even assuming that the government engaged in such “covert acts” by not revealing its role in CA’s investigation, see id.—a curious claim given that Richards was interviewed at the United States Attorney’s Office, and now concedes on appeal that he knew at his interview that he was a target of a securities fraud investigation—Richards cites no case law supporting his argument, nor can he because, as he concedes, even “[o]utside the immunity context, courts often have repeated that the Fifth Amendment grants no right to lie,” id. at 2.
While circumstances may provide a duress defense due to improper government pressure, see, e.g., United States v. Knox, 396 U.S. 77, 82-83, 90 S.Ct. 363, 24 L.Ed.2d 275 (1969), nothing about the circumstances under which Richards was interviewed rendered his false statements immune or otherwise afforded him Fifth Amendment protection, see United States v. Jacobs, 547 F.2d 772, 777 (2d Cir.1976) (“[A] constitutional claim may not be asserted as a defense to a perjury charge.”). Thus, Richards’s indictment was neither jurisdictionally deficient nor “motivated by a coerced confession,” McMann, 409 F.2d at 1023, and his coercion claim is therefore barred by his guilty plea, see Hayle v. United States, 815 F.2d 879, 881 (2d Cir.1987).
III. Do The Defendants’ Sentences Violate The Ex Post Facto Clause?
Next, Kumar and Richards contend that application of the 2005 Guidelines to their fraud offenses, which were completed in 2000, violated the Ex Post Facto clause. See U.S. Const. art. I, § 9, cl. 3; see also art. I, § 10, cl. 1 (prohibiting states from passing any ex post facto law). We believe this claim is without merit.
A “sentencing court must generally apply the version of the Guidelines that is in effect at the time of sentencing, unless there is an ex post facto problem” with such application. United States v. Rodriguez, 989 F.2d 583, 587 (2d Cir.1993) (internal citation omitted); see also U.S.S.G. § 1B1.11(a) (2008) (“The court shall use the Guidelines Manual in effect on the date that the defendant is sentenced.”). For a law to contravene the Ex Post Facto clause, “two critical elements must be present: First, the law must be retrospective, that is, it must apply to events occurring before its enactment; *625and second, it must disadvantage the offender affected by it.” Miller v. Florida, 482 U.S. 423, 430, 107 S.Ct. 2446, 96 L.Ed.2d 351 (1987) (internal quotation marks omitted); see United States v. Kilkenny, 493 F.3d 122, 126 (2d Cir.2007) (“[T]he Ex Post Facto Clause enshrines in the Constitution a basic presumption of our law, that is, legislation in the criminal law is not to be applied retroactively.” (internal quotation marks omitted)).
The defendants contend that application of the 2005 Guidelines to their fraud offenses was both “disadvantage[ous]” and “retrospective.” Miller, 482 U.S. at 430, 107 S.Ct. 2446. They argue that their fraud offenses were “completed ... in October 2000, when the New Business Model brought an end to the 35-day month practice,” Kumar Br. 16, and therefore, that application of the 2005 Guidelines to their securities fraud offenses, which substantially increased their sentences, violated the Ex Post Facto clause. The government responds by citing the “one-book rule,” in effect prior to the time of the defendants’ commission of fraud, which provides that “[i]f the defendant is convicted of two offenses, the first committed before, and the second after, a revised edition of the Guidelines Manual became effective, the revised edition of the Guidelines Manual is to be applied to both offenses.” See U.S.S.G. § 1B1.11(b)(3).
Here, there is no question that application of the 2005 Guidelines disadvantaged the defendants by subjecting them to the higher ranges of the 2005 Guidelines compared to the 1998 version of the Guidelines in effect when the 35-day month practice was discontinued in October 2000.10 Both Kumar and Richards were convicted of fraud and obstruction. In accordance with U.S.S.G. § 3D1.2(c), these crimes were grouped together for sentencing under the higher of the two applicable offense levels, which groups the separate offenses of fraud and obstruction into a single group because “one of the counts,” here, obstruction, “embodies conduct that is treated as a[n] ... adjustment to the guideline applicable to [the other] count[],” here, fraud. See U.S.S.G. § 3C1.1. The effect of the aforementioned revisions to the 1998 Guidelines, see supra n.10, resulted in a 20-level enhancement for both Kumar and Richards, from offense levels of 30, calling for a range of 97 to 121 months’ imprisonment, to offense levels of 50, calling for life imprisonment.11 The differences between *626the application of the 1998 and the 2005 version of the Guidelines is decidedly detrimental to the defendants. Thus, our only inquiry is whether application of the 2005 Guidelines to the defendants’ sentences was “retrospective” within the meaning of the Ex Post Facto clause. See Miller, 482 U.S. at 430, 107 S.Ct. 2446.12
We previously reserved ruling on this question. See United States v. Santopietro, 166 F.3d 88, 96-97 (2d Cir.1999), abrogated on other grounds, Sabri v. United States, 541 U.S. 600, 124 S.Ct. 1941, 158 L.Ed.2d 891 (2004). In Santopietro, we noted that “[t]he Commission has issued a policy statement specifying that where some offenses occur before and some occur after a revised Guidelines version, the later version is to be applied to all offenses.” Id. at 96; accord U.S.S.G. § 1B1.11(b)(3) (1998). However, the Commission’s policy statement does not mark the end of the inquiry, because “an agency’s interpretation of its own regulations” is only given “controlling weight” where its interpretation “does not violate the Constitution or a federal statute.” Stinson v. United States, 508 U.S. 36, 45, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993) (internal quotation marks omitted); accord United States v. Stephens, 347 F.3d 427, 430 (2d Cir.2003). Thus, we must independently consider whether the one-book rule violates the Ex Post Facto clause.
A majority of circuit courts has held that the one-book rule does not contravene the Ex Post Facto clause, “at least as applied ... to a series of similar offenses.” Santopietro, 166 F.3d at 96. In United States v. Vivit, the Seventh Circuit held that “the enactment of the grouping rules [under U.S.S.G. § 3D1.2] provides fair notice such that the application of §§ 1B1.11(b)(3) and 3D1.2 does not violate the Ex Post Facto Clause.” 214 F.3d 908, 919 (7th Cir.2000). According to the Seventh Circuit, the combination of the grouping rules and the one-book rule puts a defendant on notice that “the version of the sentencing guidelines in effect at the time he committed the last of a series of grouped offenses will apply to the entire group.” Id. at 918 (internal quotation marks omitted). The Fourth, Fifth, Eighth, Tenth, and Eleventh Circuits agree. See United States v. Bailey, 123 F.3d 1381, 1404-05 (11th Cir.1997) (“[T]he one book rule, together with the Guidelines grouping rules and relevant conduct, provide that related offenses committed in a series will be sentenced together under the ... Manual in effect at the end of the series.”) (footnotes omitted); see also United States v. Sullivan, 255 F.3d 1256, 1262-63 (10th Cir.2001); United States v. Lewis, 235 F.3d 215, 218 (4th *627Cir.2000); United States v. Kimler, 167 F.3d 889, 893-95 (5th Cir.1999); United States v. Cooper, 35 F.3d 1248, 1254-55 (8th Cir.1994), vacated, 514 U.S. 1094, 115 S.Ct. 1820, 131 L.Ed.2d 742 (1995), reinstated, 63 F.3d 761, 762 (8th Cir.1995) (per curiam), cert. denied, 517 U.S. 1158, 116 S.Ct. 1548, 134 L.Ed.2d 650 (1996).13
The Third and Ninth circuits, however, have rejected the Commission’s position as incompatible with the Ex Post Facto clause. In United States v. Ortland, the defendant was charged with five identical counts of mail fraud. 109 F.3d 539, 547 (9th Cir.1997). The offenses charged in four of the counts occurred prior to November 1, 1989, when the relevant Guidelines section for calculating loss in a fraud case was amended, and the fifth count covered conduct that took place in December 1989, after the amendment. Id. at 546. The district court, finding no ex post facto problem, applied the amended Guidelines section to all five counts resulting in an enhanced sentence for the first four counts. Id.
The Ninth Circuit viewed each count as a “completed” offense, and concluded that “[application of the [Commission’s] policy statement in this case would violate the Constitution; its application would cause Ortland’s sentence on earlier, completed counts to be increased by a later Guideline.” Id. at 547. The Ortland court further questioned the “logic[ ]” behind the one-book rule, opining that
[t]he harm caused by the earlier offenses can be counted in sentencing the later one. That does not mean that the punishment for the earlier offenses themselves can be increased, simply because the punishment for the later one can be. In fact, were the later count to fall at some time after sentencing, all that would remain would be the earlier sentences, which would be too long. There are, in fact, five separate crimes; each carnes its own punishment, even if the sentences are all run concurrently to the extent that they overlap.
Id. (emphasis in original) (internal citation omitted). The Ortland court effectively found that application of the one-book rule under the circumstances would be akin to the “tail wagging the dog.” See, e.g., United States v. Bertoli, 40 F.3d 1384, 1404 n. 17 (3d Cir.1994) (concluding that, “while the one-book rule ... certainly can compel application of the earlier Manual,” the Ex Post Facto clause may apply so as to prohibit the application of the later Manual to all counts); see also Santopietro, 166 F.3d at 96 (discussing the related problem, left by Ortland and Bertoli, of “whether the grouping rules of the earlier or later versions are to be applied, after each version has been used to determine the adjusted base offense level for each count”).
In Santopietro we declined to reach this issue, in part because of the circuit conflict, and in part because it was possible that the defendant’s sentence in that case would not be affected by the difference in the two Guidelines versions. Id. at 96-97.14 In this case, we must face the issue *628because the defendants’ Guidelines calculations under the later version had an unequivocally negative impact on their recommended sentences. In addition, in the decade since Santopietro was decided, courts within this circuit have repeatedly wrestled with, and are divided on, the issue. Compare United States v. Weisberg, No. 07CR66, 2008 WL 2323376, at *4-5 (W.D.N.Y. June 2, 2008) (finding no ex post facto problem where “defendant was on notice that any tax offense committed after November 2001 would face the more stringent tax table”), ivith United States v. Johnson, Nos. 97-CR-206, 98-CR-160, 1999 WL 395381, at *9-11 (N.D.N.Y. June 4, 1999) (concluding that “the most appropriate way to handle ... multiple counts in light of the [E]x [PJost [FJacto clause is to apply the earlier Sentencing Manual to those counts as to which the underlying conduct was completed before the later version became effective ... and apply the current version to counts involving subsequent conduct”). Thus, this case presents the appropriate opportunity to “provide[ ] definitive instructions on how to handle such a situation.” Johnson, 1999 WL 395381, at *10.
We conclude that the one-book rule set forth in § 1B1.11(b)(3) does not violate the Ex Post Facto clause when applied to the sentencing of offenses committed both before and after the publication of a revised version of the Guidelines. “[C]entral to the ex post facto prohibition is a concern for ‘the lack of fair notice and governmental restraint when the legislature increases punishment beyond what was prescribed when the crime was consummated.’ ” Miller, 482 U.S. at 430, 107 5.Ct. 2446 (quoting Weaver v. Graham, 450 U.S. 24, 30, 101 S.Ct. 960, 67 L.Ed.2d 17 (1981)). The existence of an ex post facto violation turns on whether an individual was deprived of fair notice, “not [on] an individual’s right to less punishment.” Weaver, 450 U.S. at 30, 101 S.Ct. 960. The Framers’ intent in requiring such notice was “to assure that legislative Acts give fair warning of their effect and permit individuals to rely on their meaning until explicitly changed.” Id. at 28-29, 101 S.Ct. 960.
Applying these principles to § 1B1.11(b)(3), we hold that the adoption of the one-book rule prior to the commission of the defendants’ obstruction offense had placed them on notice of the consequences of committing that second offense. That the consequences of the second offense included the application of the post-amendment Guidelines to all offenses considered at the defendants’ sentencing was fully apparent prior to the commission of the crimes that triggered those consequences. When the defendants committed their obstruction offenses, “it was not the amendments to the Sentencing Guidelines that disadvantaged [the defendants], it was [their] election to continue [their] criminal activity.” Cooper, 35 F.3d at 1250; accord Vivit, 214 F.3d at 919.
Our affirmance of the defendants’ sentences on this ground offends neither of the fundamental concerns—notice and governmental restraint—protected by the Ex Post Facto clause. As to notice, we observe that prior to the commission of their obstruction offenses the defendants could have altered them conduct so as to avoid any heightened punishment imposed on the basis of the one-book rule by choosing not to obstruct the government’s investigation of their prior fraud. As to governmental restraint, our holding continues to prevent the Sentencing Commission and Congress from imposing a heightened punishment following the commission of the criminal conduct triggering that punishment. As the Guidelines themselves recognize, application of the one-book rule does not, and indeed may not, entail the application of a sentencing range devised *629after the commission of all of the offenses subject to sentencing. See § 1B1.11 cmt. background (2008) (“[E]ven in a complex case involving multiple counts that occurred under several different versions of the Guidelines Manual, it will not be necessary to compare more than two manuals to determine the applicable guideline range— the manual in effect at the time the last offense of conviction was completed and the manual in effect at the time of sentencing.” (emphasis added)).15
The one-book rule, when it leads to a higher sentencing range than would be applied to a single offense, operates in a manner similar to that of the recidivist statutes and “three strikes” laws upheld by the Supreme Court and our sister circuits in the past. The Supreme Court in Gryger v. Burke, 334 U.S. 728, 68 S.Ct. 1256, 92 L.Ed. 1683 (1948), rejected the defendant’s argument that the consideration of his past offenses in determining his sentence for a later offense was foreclosed by the Ex Post Facto clause. Id. at 732, 68 S.Ct. 1256. (“[W]e [do not] think the fact that one of the convictions that entered into the calculations by which petitioner became a fourth offender occurred before the Act was passed, makes the Act invalidly retroactive.... ”). The Ninth Circuit has on several occasions upheld such laws. United States v. Ahumada-Avalos, 875 F.2d 681, 684 (9th Cir.1989) (per curiam) (upholding a repeat offender statute); see also United States v. Kaluna, 192 F.3d 1188, 1199 (9th Cir.1999) (en banc) (“The Supreme Court and this court uniformly have held that recidivist statutes do not violate the Ex Post Facto clause if they are ‘on the books at the time the [present] offense was committed.’ ” (alteration in original) (quoting Ahumada-Avalos)). The Fifth, Seventh, Eighth, and Eleventh Circuits have come to this same conclusion. See United States v. Rosario-Delgado, 198 F.3d 1354, 1356 (11th Cir.1999); United States v. Rasco, 123 F.3d 222, 227 (5th Cir.1997); United States v. Washington, 109 F.3d 335, 338 (7th Cir.1997) (“The three-strikes law was enacted before Washington committed the bank robberies, so he had fair warning of the consequences attached to new violent offenses.”); United States v. Farmer, 73 F.3d 836, 841 (8th Cir.1996).
The fact that the impetus for enacting the recidivist statutes was to reflect the greater culpability associated with the latter offenses, whereas the impetus for the enactment of the one-book rule is to avoid “piecemeal” sentencing, U.S.S.G. § 1B1.11 cmt. background (2008), is of no relevance for purposes of determining the retroactivity of the legal consequences of a defendant’s actions. It might also be argued that the recidivist statutes impose punishment upon only a single crime, the prior offenses having already been committed and for which the defendant had been sentenced. Of the seven cases cited above, however, six place no reliance on that potential distinction. Only the Eighth Circuit, in Farmer, implies that the sentencing of only the last crime that triggered the consequences of the recidivist statute was a factor supporting the constitutionality of *630the “three strikes” law. 73 F.3d at 841 (“[S]o long as the actual crime for which a defendant is being sentenced occurred after the effective date of the new statute, there is no ex post facto violation.” (quoting United States v. Allen, 886 F.2d 143, 146 (8th Cir.1989)) (internal quotation marks omitted)). Despite this statement in Farmer, the distinction between the recidivist statutes and the one-book rule makes neither a practical nor a logical difference for purposes of an analysis under the Ex Post Facto clause. In both cases, prior conduct becomes the basis for imposing a heightened sentence only upon conviction for a later criminal act.16 Here, the defendants’ obstruction offense is the “actual crime” triggering the application of the one-book rule, the defendants had pri- or notice of the consequences of that crime, and therefore the application of the one-book rule is proper.17
Kumar offers a hypothetical situation that, he claims, demonstrates the incorrectness of our ruling. Kumar Br. 25. In
*631his hypothetical, he posits two defendants, A and B, and that the fraud penalties are increased, as here, between 2000 and 2004 in the event of a conviction for both fraud and obstruction of justice. Defendant A is convicted of committing fraud in 2000 and is found at a Fatico hearing to warrant a sentencing enhancement on the basis of obstruction of justice perpetrated in 2004, although he is not tried or convicted of that crime. Defendant B is convicted of both fraud committed in 2000 and of obstructing justice in 2004. Kumar notes that in such a situation, as a result of the one-book rule, “B’s offense level [for the fraud] would be higher than A’s ... even though them conduct was the same.” Kumar Br. 25-26. The error in Kumar’s reasoning, of course, is that the hypothetical defendants face different consequences as a result of substantially different circumstances. Defendant A could have received a higher sentence for the fraud if the prosecutor had obtained a conviction for both offenses because Defendant A. would have been on notice of the legal consequences of a conviction for the obstruction, which would have triggered the one-book rule and, thus, would have included a higher sentence for fraud than would have been the case without the obstruction conviction. In this hypothetical, however, only Defendant B’s conduct resulted in a conviction for obstruction. Because the prosecutor in Defendant A’s case did not obtain a conviction for the obstruction, the sentence for the fraud is necessarily lower. The difference is not the result of an inequity in the guidelines; it is the result of a successful prosecution of Defendant B for obstruction, which triggered the operation of the one-book rule of which Defendant B had notice, and which was absent in the case of Defendant A. Kumar’s hypothetical has no ex post facto implications.
In substance, the defendants propose a rule requiring notice before the first offense of the sentencing consequences flowing from subsequent offenses. The emptiness of the notice demanded by the defendants’ position is obvious. Notice of the consequences of a subsequent offense, an offense presumably not even contemplated at the time the first offense is committed, would have no bearing on the commission of the first offense. The one-book rule only becomes relevant to a defendant at the time of the commission of a subsequent offense, when it puts the defendant on notice that when he commits that subsequent offense the sentence for his first offense will be higher than it would have been otherwise.
In rejecting the holdings of the Third and Ninth circuits, we recognize the practical—though unlikely—risk that applying the one-book rule to offenses that straddle a Guidelines revision could result in a defendant being sentenced for offenses according to the revised Guidelines only to see the second offense, which enabled the application of the one-book rule, fall out after sentencing. See Ortland, 109 F.3d at 547. The hypothetical posed in Ortland is, of course, irrelevant on the facts of this case. Kumar and Richards were sentenced based on two proper convictions, and the application of the one-book rule to their sentencings does not offend the Constitution’s Ex Post Facto clause.
IV. Was The Loss Calculation Clearly Erroneous?
Next, the defendants argue that the district court’s determination that their fraudulent conduct resulted in financial losses greater than $400 million was clearly erroneous and, as a result, improperly enhanced their Guidelines level by thirty points. The government responds that there was ample support for the district court’s calculation based on the testimony of the government’s expert at the Fatico *632hearing. We agree with the government and affirm the district court’s loss calculation.
A sentencing court is not required to compute the loss resulting from a fraud offense “with precision.” United States v. Jacobs, 117 F.3d 82, 95 (2d Cir.1997). Instead, “[t]he court need only make a reasonable estimate of the loss.” U.S.S.G. § 2B1.1, cmt. n.3 (1998). Moreover, because the district court occupies a unique position to assess the evidence and estimate the loss based upon that evidence, the court’s loss determination is entitled to appropriate deference. See United States v. Bennett, 252 F.3d 559, 565 (2d Cir.2001); United States v. Germosen, 139 F.3d 120, 129 (2d Cir.1998); cf. United States v. Rostoff, 53 F.3d 398, 407 (1st Cir.1995) (“Calculating the amount of loss for purposes of the sentencing guidelines is more an art than a science. Courts can, and frequently do, deal with rough estimates.”). While “[Bosses from causes other than the fraud must be excluded from the loss calculation,” United States v. Ebbers, 458 F.3d 110, 128 (2d Cir.2006), courts frequently calculate loss in securities fraud cases “by relying on the change of market capitalization as a result of the disclosure of the fraud,” Gov’t Br. 52, in order to prevent “perpetrators of [the] fraud” from “get[ting] a windfall,” Ebbers, 458 F.3d at 127.
The loss calculation in this case was sharply disputed. The most significant area of disagreement centered on how to properly frame the economic impact of the 35-day month practice. The government’s expert, Dr. Mukesh Bajaj, framed the loss resulting from the 35-day month practice as an “earnings miss,” which caused an estimated 10.68% decline in CA’s stock price that translated into a loss of $330 million for one quarter of fiscal year 2000 alone. Conversely, the defendants’ expert, Professor Daniel Fischel, denied that the 35-day month practice caused an “earnings miss,” whereby the earnings CA reported were completely “fabricatefd],” but instead testified that the practice only caused an “earnings shift,” whereby earnings that were properly attributable to a future quarter were reported in the previous quarter. Kumar Br. 29-30 (internal quotation marks omitted). Fischel did not submit his own loss calculation, but focused only on refuting Bajaj’s analysis.18
The district court held a Fatico hearing in order to untangle this web. During the hearing, the district court questioned both Bajaj and Fischel on their respective analyses. Specifically, the district court challenged Bajaj’s analysis as based on only one quarter’s losses, which Fischel argued resulted in an artificially inflated loss calculation. In turn, the district court questioned Fischel’s analysis insofar as seeming to imply that the fraudulent conduct caused no economic loss whatsoever. Fischel conceded, however, that, even if Bajaj’s “earnings miss” nomenclature was inaccurate, there was still an investor loss to be ascertained based not only on the market impact of the 35-day month practice on earnings, but also on the “real effect on cash flows resulting from th[e] disclosure” of the practice, which led to a drop in confidence in CA’s management and market speculation about the extent of the fraudulent activity. See Fatico Hr’g Tr. 416:13-14, Oct. 25, 2006.
Ultimately, the district court accepted Bajaj’s calculation. The court found that “[a]lthough[the] precise dollar amount of *633the loss ... is difficult if not impossible to fix, the testimony of Dr. Bajaj [was convincing] that the loss was in the hundreds of millions of dollars.” Kumar Sentencing Tr. 64:17-20. In rejecting Fischel’s analysis, the court noted that Fischel was unable to explain why CA executives would engage in the 35-day month practice if it did not, in fact, positively affect CA’s stock price. See Kumar Sentencing Tr. 64:1 (characterizing Fischel’s testimony during the Fatico hearing as “not enlightening”). Consequently, the defendants’ Guidelines calculation was enhanced thirty levels (under the 2005 Guidelines) for the financial loss.
On appeal, Kumar and Richards claim that the district court’s reliance on Bajaj’s “earnings miss” analysis was clearly erroneous, because Fischel showed that the 35-day month practice did not result in earnings misses but instead in earnings shifts whereby any earnings lost in one quarter were made up in the subsequent one. As an initial matter, the record reveals that Fischel’s objection to the term “earnings miss” was partly semantic. As the defendants now concede, Bajaj did not analyze the 35-day month practice as resulting in an earnings miss, but instead his analysis simply “removed the earnings associated with each contract from the quarter in which it had been improperly booked and ‘re-booked’ it in the proper quarter.” Kumar Br. 27. However, despite this concession, the defendants argue that Bajaj’s calculation was nevertheless erroneous because he only calculated the “miss” part of the shift, and did not take into account that, had CA “reported ... its earnings correctly, ... it would have reported higher earnings in the third and fourth quarters of fiscal year 2000 than the earnings it actually reported.” Kumar Br. 30 (emphasis in original, internal quotation marks omitted). Thus, the defendants argue that Bajaj’s calculation should have accounted for these higher earnings, which would have significantly—or completely—offset the earnings miss in the prior quarters.
We are not persuaded. Taken to its logical conclusion, the defendants’ argument would also compel the conclusion that CA’s investors did not suffer any loss due to the 35-day month practice, since under Fischel’s hypothesis, the false gains and the false losses should have effectively canceled each other out. Fischel himself undermined this conclusion during the Fatico hearing when he conceded that the 35-day month practice “obviously” had a “real negative monetary effect” on CA’s investors, not only by affecting CA’s “cash flow[ ],” but by injecting “speculation” into the market and harming investor confidence in CA’s management. Fatico Hr’g Tr. 416:13-14, 23, 419:8, 18. Accepting Fischel’s testimony would cast doubt on the entire basis for the Generally Accepted Accounting Principles rule that earnings earned in one quarter must be reported in the same quarter; if merely “shifting” earnings between quarters had no negative effect on investors, there would be no need for the rule in the first place. As Bajaj noted in his report, “[m]any firms that missed earnings in a given quarter could also have avoided announcing the miss if they could ‘borrow’ sufficient earnings from the next quarter to cover their shortfall.” Supplemental Report of Dr. Mukesh Bajaj, at 12, Oct. 9, 2006. Thus, the government properly characterizes Fischel’s analysis as “stretch[ing] credulity.” Gov’t Br. 53.
In addition, during the Fatico hearing, the government not only provided reasons why Fischel’s analysis was wrong, but also provided reasons why Bajaj’s analysis was sound. Specifically, the government showed that “re-booked” false earnings in one quarter could indeed have caused CA’s stock price to decline, by, inter alia, causing loss to investors who purchased CA *634stock at inflated prices because of fraudulent disclosures and omissions flowing from the 35-day month, then sold after such inflation had seeped out of the stock, as well as losses to investors who sold at deflated prices, where the 35-day month caused CA to understate results. See Fatico Hr’g Tr. 234-236, Oct. 25, 2006. In other words, individual sales and purchases of CA stock were singular events— thus, even assuming CA came out even in the end (a dubious contention, at best), CA’s individual investors did not go unharmed. Instead, claims filed by individual investors with CA’s claims administrator for their losses resulting from the 35-day month practice reveal that those losses not only existed, but were significantly underestimated by Bajaj, who “calculated damages only to an artificially narrow subset of victims: namely, those victims who filled out and submitted claims for restitution,” which is often “substantially fewer than all eligible victims.” Gov’t Br. 55 n.22. This potential disconnect between the financial health of a corporation and that of individual investors is precisely why the type of “event study” undertaken by Bajaj “conforms to the most widely accepted practices of economists.” Id. at 55.
The defendants also claim that Bajaj’s loss calculation was erroneous because the “sample of earnings misses” he used in his study “involved firms that had experienced genuine adverse developments ... such that a stock price decline would be expected.” Kumar Br. 30. Thus, the defendants argue that the sample firms Bajaj used in his study were, not appropriate comparators to CA. Again, the defendants’ argument is unpersuasive. In estimating a loss calculation, a sentencing court should not analyze the impact of fraud in a vacuum, but instead should recognize that “[m]any factors may cause a decline in share price between the time of the fraud and the revelation of the fraud,” not all of which will be attributable to fraudulent activity. United States v. Rutkoske, 506 F.3d 170, 179 (2d Cir.2007).
In this case, the district court properly focused on loss attributable to the defendants’ fraud. As previously established, the 35-day month practice did indeed cause a “genuine adverse development” to CA’s financial health. Moreover, Bajaj’s study took other causes for CA’s stock decline into account by comparing the earnings misses of CA to those of thirty other similarly situated companies, and “performing] a comprehensive regression analysis in an effort to isolate the event-specific impact of the given firm’s earnings-miss disclosure on its stock price, while controlling for market-wide or industry-wide factors.” Gov’t Br. 56. The district court was fully entitled to credit Bajaj’s assessment of the impact of external causes on CA’s stock and reject Fischel’s, particularly in light of Fischel’s inherently contradictory analysis.
Finally, in reviewing the district court’s loss calculation, we find it particularly telling that CA’s stock closed $5.85 below what would have been expected on October 10, 2003, two days after CA’s public disclosure of the 35-day month practice. Multiplying that undisputed figure by the 200.2 million affected shares would have brought the estimated loss to investors to more than $1 billion, alone. Even assuming that the losses of October 9 and 10, 2003 were tied, not to the actual operation of the fraudulent 35-day month practice, but to the “possibility of an- open-ended criminal and regulatory investigation that could ... [have] disastrous effects for the company,” Kumar Br. 37 (alteration in the original), financial loss caused by speculation that stems from the fraudulent practice and a loss of confidence in management is properly included in the loss calculation, see Ebbers, 458 F.3d at 127. Thus, we affirm the district court’s loss calculation of *635greater than $400 million as not clearly erroneous.
V. Were The Defendants Properly Denied Acceptance Of Responsibility Credit?
Next, the defendants argue that, “[a]lthough [they] pleaded guilty to all the charges against [them],” the district court abused its discretion in denying them “any credit for acceptance of responsibility.” Kumar Br. 44 (emphasis in original); see Richards Br. 47-48.
A defendant is entitled to a two-point reduction under U.S.S.G. § 3E1.1(a) if he “clearly demonstrates acceptance of responsibility for his offense.” A defendant “who enters a guilty plea is not entitled to an adjustment under this section as a matter of right.” U.S.S.G. § 3E1.1, cmt. n.3. In particular, a defendant who engages in “[cjonduct resulting in an enhancement under § 3C1.1 (Obstructing or Impeding the Administration of Justice) ordinarily” would not be entitled to the reduction, as such conduct “indicates that the defendant has not accepted responsibility for his criminal conduct.” Id. at cmt. n.4. However, there may be “extraordinary cases” in which a defendant who obstructs justice in some way is also entitled to a reduction for acceptance of responsibility. Id.
A district court’s § 3E1.1 determination is entitled to “great deference” because the “sentencing judge is in a unique position to evaluate a defendant’s acceptance of responsibility.” Id. at cmt. n. 5. A district court’s decision to deny credit for acceptance of responsibility, primarily a factual determination, will be upheld unless it is “without foundation.” United States v. Harris, 13 F.3d 555, 557 (2d Cir.1994).
A. Kumar’s Acceptance Of Responsibility
At sentencing, the district court denied Kumar’s request for an acceptance of responsibility reduction under § 3E1.1, concluding that Kumar had not sufficiently accepted responsibility for his criminal conduct because he had obstructed justice, and because he had waited until the eve of trial to plead guilty. In addition, the district court noted that (1) Kumar’s plea allocation was phrased to “mute[] the gravity of his complicity in the securities fraud offenses,” Sentencing Tr. 61:12-13, Nov. 2, 2006, and (2) Kumar’s meritless objections to the evidence-tampering and fraudulent transaction allegations in the PSR revealed a lack of acceptance of responsibility, Kumar Sentencing Tr. at 60:18-61:21.
On appeal, Kumar claims that the district court’s rejection of his acceptance of responsibility request was erroneous. First, Kumar claims that, although an acceptance of responsibility departure is generally unavailable when a defendant engages in obstructive behavior, that exception does not apply to him because his obstructive behavior occurred pre-indictment and the exception only applies to post-indictment obstructive behavior. See, e.g., United States v. Gregory, 315 F.3d 637, 641 (6th Cir.2003) (granting acceptance points where “[a]ll of [the defendant’s] obstructive conduct predated [the] indictment”); see also United States v. Teyer, 322 F.Supp.2d 359, 368 (S.D.N.Y.2004) (“Were courts to hold that any obstructive conduct, however early in the investigation ... of a case ... forever disentitled a defendant to credit for later acceptance of responsibility, this incentive would be ill served.”). Second, Kumar argues that the district court placed too much reliance on the “lateness” of his plea in rejecting his request for an acceptance *636of responsibility reduction, which he argues is only a basis for denying him “a third acceptance point, ... but [not] ... for denying the first two.” Kumar Br. 44; see United States v. Sloley, 464 F.3d 355, 359 (2d Cir.2006) (“[A] government motion is a necessary prerequisite to the additional one-level decrease [for a timely plea] under Guidelines § 3E1.1(b).”); see also Kumar Br. 49 (noting that “this Court has not directly addressed the relationship between the timeliness of a defendant’s guilty plea and his receipt of acceptance credit”).
We need not resolve either of these alleged flaws in the district court’s reasoning with respect to Kumar, however, because an examination of the record shows that he engaged in sufficient objectionable post-indictment conduct to justify a rejection of his request for acceptance of responsibility credit. Specifically, Kumar, individually and separately from Richards, acted in ways that the district court reasonably found to be inconsistent with a full acceptance of responsibility. For example, the district court found that “Kumar’s carefully worded plea allocution ... muted the gravity of his complicity in the securities fraud offenses.... When asked if he thought that [his fraudulent conduct] ... would have affected a prudent investor’s decision to buy or sell his company’s stock, his response was: ‘I could see the possibility where it could.’ ” Kumar Sentencing Tr. 61:11-13, 17-21. Moreover, the district court found it particularly telling that Kumar objected to the PSR on various grounds related to evidence tampering and fraudulent transactions during his tenure at CA, but then withdrew those objections during the course of the Fatico hearing, implicitly acknowledging that his objections lacked merit. Indeed, notwithstanding Kumar’s withdrawal of his objection, the district court expressly found at sentencing that there was “clear[] and convincing[]” evidence that at least one of Kumar’s objections was meritless. Kumar Sentencing Tr. 61:9. See U.S.S.G. § 3E1.1, cmt. n.1(a) (“[A] defendant who falsely denies, or frivolously contests, relevant conduct that the court determines to be true has acted in a manner inconsistent with acceptance of responsibility.”).
Thus, Kumar’s post-indictment conduct provided a reasonable basis for rejecting his request for an acceptance of responsibility reduction.
B. Richards’s Acceptance Of Responsibility
While the district court set forth several reasons why Kumar did not deserve a sentencing reduction for acceptance of responsibility, in denying Richards’s same request, the court relied on a single factor: the lateness of Richards’s plea. According to the district court, “the most significant factor in the acceptance of responsibility scale is the factor of time limits.” Richards Sentencing Tr. 15:7-9. The district court concluded that, by pleading two weeks before trial, Richards had exceeded those “time limits,” and therefore, was not entitled to acceptance of responsibility credit. On this point, we disagree.
Timeliness of a defendant’s plea is an appropriate consideration in the acceptance of responsibility determination. U.S.S.G. § 3E1.1 cmt. n.1(h). However, while the two-level reduction provided for in § 3E1.1(a) is for demonstration of acceptance of responsibility, the Sentencing Guidelines specifically provide that timeliness of a plea is primarily relevant to the reduction of an additional point under § 3E1.1(b), ostensibly “for helping the authorities save resources.” United States v. Ortiz-Torres, 449 F.3d 61, 76 (1st Cir.2006) (internal quotation marks omitted). See United States v. Eyler, 67 F.3d 1386, 1390-91 (9th Cir.1995) (“While the key inquiry for purposes of [§ 3E1.1(a)] is *637whether the defendant has demonstrated contrition, once this has been determined, then the focus of the section (b) inquiry is on timeliness.” (emphasis in original)). As Judge Lynch stated in Teyer:
The Guidelines specifically provide a sanction for belated guilty pleas that fail adequately to save the resources of the Government and the Court. Defendants who plead at an early stage qualify for an additional one-level reduction that is correspondingly unavailable for belated pleas. That a defendant can earn an additional one-level reduction merely by notifying the authorities of his intention to plead early enough to permit the government to avoid preparing for trial strongly implies that defendants may qualify for the basic two-level adjustment for acceptance of responsibility without doing so.
322 F.Supp.2d at 376 (alteration, internal citations, and quotation marks omitted).
Here, the lateness of Richards’s plea on its own was not a sufficient foundation for denying him any acceptance of responsibility credit. While under certain circumstances the lateness of a plea might indeed weigh against the defendant, those circumstances are not present in this case. For example, Richards’s plea came neither “after the [gjovernment ... concluded presenting its case, ... during the jury’s deliberations,” nor “on the morning of trial,” and the district court cited no “other [relevant] factors” that warranted a rejection of Richards’s request. Id. It is undisputed that Richards’s obstruction and culpability was “of a different order” than that of Kumar. Richards Sentencing Tr. 25:15-16. Unlike Kumar, Richards appeared to fully accept responsibility both prior to and during sentencing. See Richards Sentencing Tr. 12:14-16 (‘Your Honor, I fully accept responsibility for the actions that I have taken and regret those actions.”). Also unlike Kumar, who tampered with physical evidence, bribed witnesses, and lied repeatedly during the course of a federal investigation, Richards’s obstructive conduct consisted entirely of a single, albeit false, denial of knowledge of the 35-day month practice during an interview and subsequent testimony that predated his guilty plea by three years and was accounted for in the indictment. Indeed, denying Richards acceptance of responsibility solely due to his charged obstructive conduct would effectively raise the Guidelines level for such conduct by foreclosing the opportunity for acceptance of responsibility credit in obstruction cases.
Because “the paramount factor in determining eligibility for § 3E1.1 credit is whether the defendant truthfully admits the conduct comprising the offense or offenses of conviction,” and because Richards did so here “in a sufficiently timely manner so as to avoid a lengthy trial and avert the risk of an irrational verdict,” Teyer, 322 F.Supp.2d at 376, his lateness in pleading guilty did not by itself provide a sufficient basis for the denial of more than a one-point reduction for acceptance of responsibility. Accordingly, because the district court erroneously failed to award Richards a two-point reduction for acceptance of responsibility that he should have received, Richards’s sentence is vacated and the case remanded to the district court for resentencing on this additional basis.
VI. Was Richards’s Sentence Substantively Unreasonable?
Finally, Richards contends that his non-Guidelines sentence of seven years’ imprisonment was unreasonably long, despite the PSR’s recommendation of a life sentence. However, because we vacate Richards’s sentence as procedurally unsound and remand the case for resentencing, there is no need to entertain his substantive reasonableness argument at this time. See Gall v. United States, 552 U.S. 38, 128 S.Ct. 586, 597, 169 L.Ed.2d 445 (2007) (“Assum*638ing that the district court’s sentencing decision is procedurally sound, the appellate court should then consider the substantive reasonableness of the sentence imposed under an abuse-of-discretion standard”) (emphasis added); accord United States v. Cavera, 550 F.3d 180, 190 (2d Cir.2008) (en banc).
CONCLUSION
For the foregoing reasons, the district court’s judgment and sentence as to Kumar is AFFIRMED in all respects; the district court’s judgment as to Richards is AFFIRMED, but Richards’s sentence is VACATED and REMANDED to the district court for resentencing consistent with this opinion.