Opinion for the Court filed by Circuit Judge TATEL.
Appellant challenges his conviction and sentence on drug and gun offenses, arguing (1) that the district court’s use of compound voir dire questions prevented him from learning about possible juror bias; (2) that the district court committed multiple errors in determining his sentence; and (3) that he received ineffective assistance of counsel at trial. Although we have repeatedly expressed our concerns about compound voir dire questions, in this case we are limited to reviewing the district court’s actions for plain error, a showing that appellant fails to make. Nor have we any basis for vacating the sentence: appellant’s Apprendi claim fails under plain error review, the sentence is reasonable, and appellant points to no evidence that the district court misunderstood its sentencing authority. In keeping with our general practice, however, we remand to the district court for an evidentiary hearing on appellant’s ineffective assistance of counsel claims because the trial record does not conclusively show whether appellant is entitled to relief.
I.
The case against appellant Willie Moul-ing stems from cocaine and a handgun found in a parka abandoned by a suspect who fled from police after having been stopped in connection with a hit-and-run accident. Although never charged with the hit-and-run that originally precipitated the investigation, Mouling was charged with and tried for possession of cocaine base with intent to distribute, 21 U.S.C. § 841(a)(1), (b)(l)(iii), using or carrying a firearm during a drug-trafficking offense, 18 U.S.C. § 924(c), and unlawful possession of a firearm and ammunition by a convicted felon, 18 U.S.C. § 922(g)(1). At trial Mouling’s defense centered on a theory of mistaken identity, namely that the police chased a different individual, the owner of the drug- and gun-containing parka, and ended up arresting Mouling instead.
Events leading up to the chase began when D.C. Metropolitan Police Department Officer Seth Anderson responded to the hit-and-run report and interviewed a witness who described the driver as a black male with a slim build, wearing black pants and a black parka with gray fur around the hood. Canvassing the area, Anderson saw an individual matching this description climbing into a parked blue Isuzu SUV. Anderson blocked the SUV *661with his squad car and questioned the driver. At trial Anderson testified that his encounter with the suspect lasted one to one and a half minutes. Anderson further testified that the individual produced a Virginia driver’s license bearing the name Willie Mouling, though a defense witness testified that the person he saw talking to an officer next to the SUV was not Moul-ing, but rather the owner of the SUV, whom the witness had regularly seen around the neighborhood. Other defense evidence indicated that Mouling drove an Accord, not an SUV.
When Anderson told the suspect that he was investigating a hit-and-run, the suspect became nervous and began reaching into his pockets. Instructed by Anderson to remove his hands from his pockets, the individual fled on foot, managing to slip out of his parka when Anderson tried to grab him. Dropping the parka, Anderson gave chase. The path of the chase was disputed at trial, with Anderson’s description of the route differing somewhat from another officer’s and from measurements of the area taken by a defense investigator indicating that the path Anderson described was actually blocked by a fence. According to Anderson, he never lost sight of the suspect and remained within fifteen feet of him throughout the chase, which he said lasted less than a minute. In the end Anderson arrested Mouling in an alley behind a neighboring street.
Returning to the vehicles, Anderson retrieved the abandoned parka and discovered a loaded handgun inside. Police also found three “cookies” — two of a white substance and one of a cream-colored substance — in the coat, each in its own plastic baggie. Neither the gun nor the bags yielded usable fingerprints.
Anderson testified that Mouling twice signaled his ownership of the parka by referring to it as “my” coat and by stating in regard to the charges he was facing, “well, you know what’s in the coat.” On cross-examination, however, Anderson acknowledged that when Mouling first saw the parka after his arrest, he denied it was his. Two of Mouling’s neighbors testified they saw him that day wearing a black quilted jacket with no hood, although they acknowledged they had no idea how many coats Mouling owned. Mouling’s booking photo showed him wearing a black quilted jacket with a collar but apparently without a hood, and the inventory of his clothing listed a black “jacket,” which the government suggests he could have obtained from a family member or a “sympathetic police officer,” Appellee’s Br. 22.
The jury convicted Mouling on all three counts. Given Mouling’s criminal history category of IV, the presentence report proposed a sentencing guidelines range of 168-210 months for drug possession based on a drug quantity of 50-150 grams of cocaine base, plus a 60-month mandatory sentence for using or carrying a firearm. The government requested a 228-month sentence, which reflected the low end of the guidelines range. Because Mouling’s trial counsel died in a car accident before sentencing, replacement counsel represented Mouling at sentencing.
The trial court sentenced Mouling to 228 months: 168 months for drug possession and 120 months for gun possession to be served concurrently, and a consecutive 60-month sentence for using or carrying a firearm during a drug-trafficking offense. The trial court also ordered concurrent terms of supervised release: five years for drug possession, three years for firearm use, and two years for gun possession.
Mouling appeals, objecting to the court’s conduct of voir dire in selecting his jury, challenging several aspects of his sentencing, and arguing that he received ineffec*662tive assistance of counsel at trial. We address each challenge in turn.
II.
We begin with Mouling’s challenge to the district court’s use of compound voir dire questions. Because we have reviewed this particular district court’s voir dire questioning multiple times, we offer only a brief description of the practice. As we explained in United States v. West, 458 F.3d 1 (D.C.Cir.2006), United States v. Littlejohn, 489 F.3d 1335 (D.C.Cir.2007), and United States v. Harris, 515 F.3d 1307 (D.C.Cir.2008), the district court’s practice was to ask potential jurors several two-part questions, instructing them to listen to both parts of the question before responding. The first part of the question asked whether jurors had a certain background characteristic or experience, and the second part asked whether in light of that characteristic or experience they thought they would have trouble being impartial. Only if a potential juror would answer “yes” to both parts of the question was she to raise her hand in response. If the answer to either part of the question was “no,” the potential juror wasn’t to respond at all. For example, the first part of one question asked whether any potential juror or any close family member or friend was “currently or previously employed by any law enforcement agency.” Trial Tr. at 58 (Sept. 21, 2004). The district court then listed various organizations that it said qualify as law enforcement agencies, warned the potential jurors not to raise their hands until it asked the second part of the question, and then asked: “As a result of that experience, do you believe that you, you personally would be unable to be fair and impartial to both sides if selected as a juror in this case?” Id. at 58-59. In addition to the law enforcement employment question, the district court posed compound questions on seven other topics: whether any prospective jurors knew each other or had been involved in criminal defense, studied law, served on a grand jury, served on a petit criminal jury, participated in a crime-prevention group, or had been the victim of any crime.
We have previously expressed “deep reservations about [the district court’s] compound questions.” Littlejohn, 489 F.3d at 1343. As we explained in West, the problem with compound questions is that they “prevent! ] the parties from learning the factual premise of the first part of the question, relying instead upon the juror’s self-assessment of his or her impartiality.” 458 F.3d at 10-11. Here, for example, if a potential juror had actually been employed by a law enforcement agency but thought she could nonetheless be impartial, the question format would prevent the parties from learning about and inquiring into the juror’s law enforcement background altogether.
In all three of our prior cases, because defense counsel timely objected to the compound questions, we reviewed the conduct of voir dire for abuse of discretion, explaining that reversal was warranted if the court abused its discretion and there was substantial prejudice to the accused. See, e.g., Littlejohn, 489 F.3d at 1342. In West and Hams, although we found the compound questions “troubling” and cautioned against their use, Harris, 515 F.3d at 1311, we nonetheless saw no abuse of discretion because the defendants had other means to learn the necessary information about potential jurors, because their cases did not turn on police officer credibility, and because the evidence against them was otherwise strong. Id. at 1313; West, 458 F.3d at 8-9. By contrast, in Littlejohn, where police officer credibility was central to conviction and the evidence *663of guilt was otherwise not overwhelming, we concluded that the compound questions violated the defendant’s Sixth Amendment right to an impartial jury and vacated the conviction. 489 F.3d at 1346.
Unlike in Harris, West, and Little-john, Mouling’s trial lawyer failed to object at voir dire to the compound questions, so our review is far more limited. See United States v. Caldwell, 543 F.2d 1333, 1345 (D.C.Cir.1975); Fed.R.CRImP. 52(b). Under plain error review, we may reverse only if: “(1) there is error (2) that is plain and (3) that affects substantial rights, and (4) we find that the error ‘seriously affects the fairness, integrity or public reputation of judicial proceedings.’ ” United States v. Baugham, 449 F.3d 167, 183 (2006) (quoting United States v. Olano, 507 U.S. 725, 732, 113 S.Ct. 1770, 123 L.Ed.2d 508 (1993)).
Mouling argues that his case resembles Littlejohn, where we held that the compound questions posed in that case violated the Sixth Amendment. According to Mouling, the law is therefore crystal clear, and the court committed plain error when it employed such questions in empaneling his jury. As the government points out, however, Littlejohn had not been decided at the time of Mouling’s trial. According to the government, any error in using compound questions could therefore not have been “plain.” In response, Mouling cites Johnson v. United States, 520 U.S. 461, 467-68, 117 S.Ct. 1544, 137 L.Ed.2d 718 (1997), for the proposition that the plainness of an error is assessed as of the time of appeal, not as of the time of trial. Since Littlejohn was issued prior to this appeal, Mouling insists that the district court’s error was plain.
Mouling’s reliance on Johnson is misplaced. In Johnson “the law at the time of trial was settled and clearly contrary to the law at the time of appeal.” 520 U.S. at 468, 117 S.Ct. 1544. In such circumstances, the Court explained, “it is enough that an error be ‘plain’ at the time of appellate consideration.” Id. Here, by contrast, the law was far from settled at the time of trial. We had yet to decide West, Littlejohn, or Harris, and no other D.C. Circuit precedent had clearly held that compound voir dire questions constitute reversible error.
In circumstances like those we face here — where the law is unsettled at the time of trial but settled at the time of appeal — whether we assess error as of the time of trial or the time of appeal remains an open question in this circuit. Baugham, 449 F.3d at 183 (recognizing that the Supreme Court has left the question “unresolved” but not deciding the question for this circuit). We have twice declined to reach the question because in those cases, even assuming any error was plain, the defendants failed to show that the error affected their substantial rights as required under the plain error test’s third element. See id.; United States v. Johnson, 437 F.3d 69, 74 (D.C.Cir.2006). Mouling, however, presents a closer case. Like in Littlejohn, officer credibility was crucial to Mouling’s conviction. The evidence on the central question — whether Mouling was the individual stopped by Officer Anderson and thus the owner of the parka — came primarily, indeed almost entirely, from police officer testimony. Officer Anderson identified Mouling as the SUV driver he stopped for questioning and testified that he never lost sight of the suspect during the chase and that Mouling made comments indicating he owned the parka. That testimony was contradicted not only by a defense witness who saw Anderson speaking to someone other than Mouling at the SUV and by defense evidence indicating the chase route was blocked by a fence, but also by another *664officer whose description of the chase route differed from Anderson’s. To convict Mouling, then, jurors had to credit Officer Anderson’s version of events, making it all the more important for voir dire to uncover any juror’s tendency to give undue weight to officer testimony. Because of the similarities between Mouling’s case and Littlejohn, where we found a constitutional violation, whether the error affected Mouling’s substantial rights presents a close question. As a result, we must now resolve the question we have previously been able to avoid: whether error can be “plain” when the law, though unsettled at trial, becomes clear by the time of appeal.
After the Supreme Court issued its decision in Johnson, the circuits have split on this question. See generally HaRry T. Edwards & Linda A. Elliott, Federal Courts Standards of Review: Appellate Court Review of District Court Deoisions and Agenoy Aotions 92 (2007). The Eleventh assesses error as of the time of appeal, United States v. Smith, 459 F.3d 1276, 1283 (11th Cir.2006), while the Ninth does so as of the time of trial, United States v. Turman, 122 F.3d 1167, 1170 (9th Cir.1997). We agree with the Ninth Circuit that Johnson represents an exception to the general rule that error is assessed as of the time of trial, an exception Johnson carved out because when the law is settled at the time of trial, “objections are pointless,” id. at 1170, and “[mjeasuring error at the time of trial ‘would result in counsel’s inevitably making a long and virtually useless laundry list of objections to rulings that were plainly supported by existing precedent,’ ” id. (quoting Johnson, 520 U.S. at 468, 117 S.Ct. 1544). By contrast, where the law is unsettled at trial, objections are far from pointless — they serve a valuable function, alerting the district court to potential error at a moment when the court can take remedial action. Thus the interest in requiring parties to present their objections to the trial court, which underlies plain error review, applies with full force. We therefore hold that where, as here, the law was unsettled at the time of trial but became settled by the time of appeal, the general rule applies, and we assess error as of the time of trial.
Because at the time of Mouling’s trial, no clear circuit precedent established the impropriety of compound voir dire questions in circumstances similar to Mouling’s case, any error in employing such questions cannot have been plain. See United States v. Perry, 479 F.3d 885, 893 n. 8 (D.C.Cir.2007) (noting that “absent precedent from either the Supreme Court or this court, [an] asserted error falls far short of plain error” unless it violates a legal norm that is “absolutely clear (for example because of the clarity of a statutory provision or court rule)” (internal quotation marks and ellipses omitted)). We therefore have no need to reach the plain error test’s remaining two elements: whether the voir dire affected Mouling’s substantial rights or “seriously affeet[ed] the fairness, integrity or public reputation of judicial proceedings,” Olano, 507 U.S. at 732, 113 S.Ct. 1770 (internal quotation marks omitted).
III.
Appealing his sentence, Mouling argues that the district court committed Apprendi error when it sentenced him without a jury finding on the requisite quantity of drugs; that it based its sentence on an unreasonable rationale; and that it failed to recognize its authority to consider the sentencing guidelines’ disparity between crack and powder cocaine.
We start with Mouling’s argument that the district court erred in sentencing him based on a drug quantity that the jury *665never found beyond a reasonable doubt. Because Mouling’s counsel failed to object on this ground at sentencing, our review is once again limited to plain error. See United States v. Johnson, 331 F.3d 962, 964 (D.C.Cir.2003).
Under Apprendi v. New Jersey, the jury, not the court, must find any facts “that increase the prescribed range of penalties to which a criminal defendant is exposed.” 530 U.S. 466, 490, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000) (internal quotation marks omitted). Such facts include drug quantity under 21 U.S.C. § 841(b)(1), which provides different mandatory sentence ranges based on the quantity of drugs involved. United States v. Fields, 242 F.3d 393, 395 (D.C.Cir.2001). Under section 841(b)(1)(A), possession of fifty grams or more of cocaine base carries a penalty range of ten years to life; under section 841(b)(1)(B), possession of five grams or more carries a range of five to forty years; and under section 841(b)(1)(C), possession of any amount carries a penalty of up to twenty years.
Mouling argues that because the court instructed the jury only that it had to find beyond a reasonable doubt that he possessed a “detectable amount” of cocaine base, Trial Tr. at 38 (Sept. 28, 2004), and omitted any reference to “50 grams or more,” it should have sentenced him under subsection (C) (detectable amount) rather than under subsection (A) (fifty grams or more). The government points out that although the jury instructions referred only to a “detectable amount,” the verdict form included the necessary quantity. Therefore, according to the government, by checking “guilty” on the verdict form, the jury actually found that Mouling possessed fifty grams or more, eliminating any Apprendi error. We disagree.
Including the quantity in the verdict form cannot cure the omission from the jury instructions. We presume that juries follow the instructions they are given, see Richardson v. Marsh, 481 U.S. 200, 211, 107 S.Ct. 1702, 95 L.Ed.2d 176 (1987), and here, the instructions mentioned only a “detectable amount.” Properly following instructions, the jury would have been required to find Mouling guilty if it concluded that he possessed only a detectable amount, regardless of what the verdict form said. The court therefore committed error in sentencing Mouling under subsection (A), and under Apprendi and Fields such error was plain at the time of trial.
Moving on, then, to the remaining elements of the plain error inquiry, we must consider whether the error affected substantial rights and “seriously affect[ed] the fairness, integrity or public reputation of judicial proceedings.” Olano, 507 U.S. at 732, 113 S.Ct. 1770 (internal quotations omitted). In United States v. Webb, the district court had improperly imposed a sentence under subsection (A), but we nonetheless held that this Apprendi error did not affect substantial rights because the actual sentence, like Mouling’s, fell below the statutory maximum available under subsection (C). 255 F.3d 890, 898 (D.C.Cir.2001). Mouling argues that Webb no longer controls because United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005), has made the sentencing guidelines advisory. According to Mouling, under a mandatory guidelines regime, there is no effect on substantial rights because the defendant would have received the same sentence based on the guidelines’ drug quantity table under either subsection (A) or subsection (C), but this is no longer true under the advisory guidelines. We needn’t decide whether Booker changes our analysis under the plain error test’s third element, however, because Mouling cannot show that his sentence meets the test’s fourth element— *666that the error “seriously affects the fairness, integrity or public reputation of judicial proceedings,” Olano, 507 U.S. at 732, 113 S.Ct. 1770 (internal quotation marks omitted).
In Webb, we held that a sentence based on an Apprendi error did not satisfy the plain error test’s fourth element when the evidence of the higher, subsection (A) drug quantity was overwhelming and uncontro-verted. 255 F.3d at 901-02; see also United States v. Cotton, 535 U.S. 625, 633-34, 122 S.Ct. 1781, 152 L.Ed.2d 860 (2002) (no plain error in failure to include drug quantity in indictment where evidence was “overwhelming” and “essentially uncontro-verted”). Attempting to distinguish Webb, Mouling (1) questions the strength of the evidence of drug quantity and (2) suggests that the difference in color among the three “cookies” found in the parka could create a reasonable doubt that all three cookies contained cocaine base, which could reduce the total weight to below the fifty-gram mark.
As to the first point, Mouling acknowledges that the chemist’s report found a total weight of 51.7 grams, but argues that the record contains no evidence as to the accuracy of the chemist’s measurements and points out that those measurements were contradicted by the government’s own expert. He also emphasizes that 51.7 grams is close to the fifty-gram breakpoint between subsection (A) and subsection (B), rendering more significant any doubt as to the accuracy of the weighing method. Mouling’s efforts to undermine the strength of the evidence are unpersuasive. Although it is true that the government never demonstrated the accuracy of the chemist’s measurements, nothing in the record suggests that the measurements were in any way inaccurate. And although the government expert estimated that the baggies represented “three half-ounce bags,” Trial Tr. at 61 (Sept. 23, 2004), this estimate, far from contradicting the chemist’s measurements (as Mouling suggests), was itself based on those measurements. See Trial Tr. at 59 (Sept. 23, 2004) (expert testimony noting the measured weight of 51.7 grams and concluding that each bag contained “approximately 17 grams each, which is just slightly over a half an ounce on each bag”). To be sure, this case involves quantities close to the fifty-gram mark, but the amount of drugs recovered need not vastly exceed the statutory amount for evidence of quantity to be “overwhelming.” The government produced physical evidence — the actual cookies themselves — and presented laboratory analyses to establish quantity. “This was not a case, for example, in which the government recovered a quantity of drugs less than the 50-gram statutory threshold, and thus had to rely on ‘vague testimonial’ rather than physical evidence to prove that the threshold was met.” United States v. Pettigrew, 346 F.3d 1139, 1147 (D.C.Cir.2003) (quoting United States v. Fields, 251 F.3d 1041, 1045 (D.C.Cir.2001)).
As to Mouling’s second point, despite his efforts on appeal, the evidence of drug quantity, as in Webb, was “essentially un-controverted,” Cotton, 535 U.S. at 633, 122 S.Ct. 1781. See also Johnson, 331 F.3d at 969 (finding failure to submit drug quantity to jury was not plain error when defendant “offered the jurors no scenario under which they could have convicted him of unlawful possession with intent to distribute cocaine base, yet found that the quantity involved was less than 50 grams”). Neither Mouling’s trial counsel nor the lawyer who represented him at sentencing questioned the validity of the weight listed in the chemist’s report or in any other way suggested that Mouling might actually have possessed less than fifty grams. Nor does Mouling now give us any reason to doubt the chemist’s measurements. For *667the first time, Mouling suggests an alternative “fake cookie” theory, which he contends could have created reasonable doubt as to the total weight of cocaine base. Mouling suggests that because the chemist tested samples from all three cookies mixed together, the test couldn’t determine whether each individual cookie contained cocaine base. If only the two white cookies, but not the cream-colored cookie, actually contained cocaine, the total weight would have fallen within the range applicable to subsection (B) (five grams or more) rather than exceeding fifty grams under subsection (A). To be sure, we have suggested that presenting on appeal a plausible scenario under which a jury might have convicted the defendant but still found he possessed less than the requisite quantity of drugs could inform the analysis under the plain error test’s fourth element. See Webb, 255 F.3d at 902. But Mouling presented neither evidence nor argument at trial, sentencing, or on appeal that the difference in cookie color on which his theory depends is in fact linked to the presence or absence of cocaine. Indeed, despite the color difference, the government’s expert testified that the three cookies “appear[ ] to be ... crack cocaine.” Trial Tr. at 57 (Sept. 23, 2004). Mouling’s unsubstantiated theory, presented for the first time on appeal, is thus insufficient to seriously affect the fairness, integrity, or public reputation of judicial proceedings.
Mouling makes an additional related argument: that the same Apprendi error requires remand for resentencing on the term of supervised release. In imposing a five-year term of supervised release, the district court expressly tracked the mandatory minimum under section 841(b)(1)(A). Section 841(b)(1)(C), under which Mouling should have been sentenced, carries no mandatory minimum. In United States v. Graham, we held on plain error review that this exact error affected the defendant’s substantial rights, even though a five-year term would have been permissible, albeit not mandatory, under section 841(b)(1)(C). 317 F.3d 262, 273-75 (D.C.Cir.2003). Although remanding for resentencing on the term of supervised release, Graham apparently never considered the plain error test’s fourth element, which, as demonstrated above, is fatal to Mouling’s Apprendi claim. Moul-ing’s challenge to the supervised release term therefore fails for the same reason as his challenge to the term of imprisonment.
Mouling next objects to the way in which the district court considered his decision to proceed to trial rather than accept a plea deal. Starting from the premise that the drug charge carried a minimum of ten years and the firearm charge a minimum of five years, the district court reasoned that Mouling should receive a heftier sentence than the fifteen years he would have received had he pled guilty. When the government pointed out that it had actually offered Mouling a substantially lower sentence, the district court declined to inquire into the details of the actual deal, stating, “Well, of course, the court doesn’t get in the middle of the plea negotiation process.” Sent’g Tr. at 10 (Oct. 31, 2005).
Although the district court declined to consider the actual plea deal the government originally offered, Mouling has given us no basis for concluding that the court acted unreasonably in focusing instead on the sentence Mouling would have gotten had he pled guilty to all charges for which he was eventually convicted. We read the district court merely to have recognized that a fifteen-year sentence would have been at the low end of the guidelines range Mouling would have received had he been eligible for the acceptance of responsibility adjustment and to have denied Mouling *668the benefit of that adjustment. The district court’s decision to impose a within-guidelines sentence absent acceptance of responsibility was reasonable. Although the district court wouldn’t have erred had it considered the actual plea deal, it was not required to do so, nor did it base the sentence on any clearly erroneous factual findings. Given that the district court correctly calculated Mouling’s guidelines range, treated the guidelines as advisory, considered the required 18 U.S.C. § 3553(a) factors, and explained its reasoning adequately to permit appellate review, it committed no procedural error. See Gall v. United States, — U.S. -, 128 S.Ct. 586, 597-98, 169 L.Ed.2d 445 (2007).
Finally, Mouling urges us to remand for resentencing in light of Kimbrough v. United States, which confirms that under the advisory guidelines, a district court may sentence below the applicable guidelines range in order to account for the sentencing disparity between powder and crack cocaine, — U.S. -, 128 S.Ct. 558, 575, 169 L.Ed.2d 481 (2007). At sentencing, Mouling’s counsel urged the court to consider this disparity and reduce Mouling’s sentence accordingly. Given the district court’s failure to expressly address this point, Mouling thinks that the court may have misunderstood its authority to consider the disparity and that remand is necessary to allow the court to do so now.
We generally remand for reconsideration only upon some record showing that the district court misunderstood its sentencing authority. See United States v. Godines, 433 F.3d 68, 70 (D.C.Cir.2006) (refusing to remand where nothing in the record rebutted “the presumption ‘that the district court knew and applied the law correctly.’ ” (quoting United States v. Ayers, 428 F.3d 312, 315 (D.C.Cir.2005))). Here, Mouling’s counsel strenuously argued that the court should consider the crack-powder disparity. Although the government insisted that the court lacked authority to do so, the record contains no indication that the district court agreed with the government or otherwise misapprehended its authority. To the contrary, the court repeatedly noted the advisory nature of the guidelines. Although Moul-ing now argues that a presumption that the district court knew and applied the law correctly is unjustified where, as here, the court was silent as to its view on the matter and the law is “completely up in the air,” Appellant’s Reply Br. 20, Mouling concedes that at the time of sentencing in this case some district courts were permissibly considering the disparity. The law therefore was hardly so inscrutable that a district court, particularly one that clearly understood the advisory nature of the guidelines, could not be presumed to follow it correctly. Remand on this ground is therefore unwarranted.
IV.
This brings us finally to Mouling’s ineffective assistance of counsel claim. For Mouling to succeed, he “must show two things: that his lawyer made errors ‘so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment,’ and that counsel’s deficient performance was prejudicial, ie., that there is a ‘reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.’ ” United States v. Gaviria, 116 F.3d 1498, 1512 (D.C.Cir.1997) (citation omitted) (quoting Strickland v. Washington, 466 U.S. 668, 687, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984)).
In this circuit, when an appellant makes an ineffective assistance of counsel claim for the first time on appeal, we generally remand for “a fact-finding hearing, at which the district court can explore *669‘whether alleged episodes of substandard representation reflect the trial counsel’s informed tactical choice or a decision undertaken out of ignorance of the relevant law.’ ” United States v. Fennell, 53 F.3d 1296, 1303 (D.C.Cir.1995) (quoting United States v. Cyrus, 890 F.2d 1245, 1247 (D.C.Cir.1989)), modified on reh’g, 77 F.3d 510 (D.C.Cir.1996). We have recognized two exceptions to this general practice: “when the trial record alone conclusively shows that the defendant is entitled to no relief,” and the “rare exception when the trial record conclusively shows the contrary.” Fennell, 53 F.3d at 1303-04.
Mouling makes six arguments in support of his ineffective assistance of counsel claim: that defense counsel (1) failed to advise him of the government’s offer of a five-year sentence in exchange for a guilty plea; (2) failed to object to the compound questions at voir dire; (3) unnecessarily introduced a police report containing prejudicial hearsay and failed to call two defense witnesses who would have testified about the identity of the hit-and-run suspect; (4) misled him into believing that he was bound by his earlier decision not to testify; (5) failed to renew the motion for judgment of acquittal despite the insufficient evidence of drug quantity; and (6) failed to make adequate inquiries into whether jury deliberations were affected by a juror who asked to be excused due to her belief that she observed defense counsel improperly communicating with a witness.
We need look no further than Mouling’s first allegation' — 'that his counsel failed to advise him of the government’s plea offer' — to agree that remand is necessary. In United States v. Gaviria, we remanded for an evidentiary hearing when defense counsel gave the defendant incorrect information about the length of sentence offered by the government in exchange for a guilty plea. 116 F.3d at 1512, 1514. We indicated that the hearing-should address whether the defendant “would have taken the Government’s plea offer had he known of his true [sentencing] exposure.” Id. at 1514. Similarly, Mouling claims that his counsel informed him only of a fifteen-year plea offer, although the government had also offered a five-year deal. The government argues that Mouling cannot have been prejudiced by counsel’s failure to inform him of the deal because he learned of the plea offer when the prosecutor mentioned it on the record in his presence. But the prosecutor’s on-the-record reference to the offer identified only the statutory provision to which the government would have allowed Mouling to plead, not the sentencing exposure that would have flowed from such a plea. As in Gaviria, then, an evidentiary hearing is needed to evaluate whether Mouling knew the details of the plea offer and whether there was a reasonable probability that he would have accepted the offer had counsel properly informed him of it.
Given that we must remand for an evi-dentiary hearing on this claim, we shall also leave it to the district court to consider Mouling’s other ineffective-assistance allegations in the first instance.
V.
While otherwise rejecting Mouling’s challenges, we remand to the district court for an evidentiary hearing on his ineffective assistance of counsel claims.
So ordered.