For a case to be within the diversity jurisdiction of the federal courts, diversity of citizenship must be “complete,” meaning that no plaintiff may be a citizen of the same state as any defendant. Straw-bridge v. Curtiss, 7 U.S. (3 Cranch) 267, 2 L.Ed. 435 (1806). The formal designation of a party in the complaint as plaintiff or defendant is not controlling; “the court will look beyond the pleadings, and arrange the parties according to their sides in the dispute.” City of Dawson v. Columbia Avenue Saving Fund, Safe Deposit, Title, & Trust Co., 197 U.S. 178, 180, 25 S.Ct. 420, 421, 49 L.Ed. 713 (1905). Though the issue has not *1265been raised, we must consider on our own initiative whether one of the nominal defendants in this case, Scotty Smith Construction Company (“Scotty”), should be realigned on the side of the plaintiff, the Fidelity and Deposit Company of Maryland, thus destroying complete diversity. Only if we decide that realignment is not required may we proceed to the substantive issues on this appeal, which are issues of Wisconsin contract law.
Two Wisconsin towns got together and hired Scotty, a Wisconsin corporation, to build for $710,000 an incinerator for burning their garbage. The towns required Scotty to post a performance bond in that amount. Fidelity, a citizen of Maryland for diversity purposes, see 28 U.S.C. § 1332(c), was the surety on the bond, and required Scotty to agree to indemnify it should it be forced to make good on the bond. Scotty built the incinerator, but the incinerator’s air pollution control device — a gas scrubber that had been supplied by one subcontractor, Krebs Engineers, a California corporation, and installed by another, Midwesco, Inc., an Illinois corporation — did not perform up to the standard required by law and the incinerator had to be shut down. The towns took the position that the scrubber’s failure to perform up to standard was a breach of contract by Scotty, and as they had paid all but $38,000 of the purchase price for a facility they could not use at all, they notified Fidelity that it must make good the difference in accordance with the bond.
Fidelity responded, several months later, by bringing this diversity action against the towns, Scotty, and the subcontractors. The complaint sought a declaration that Fidelity was not liable to the towns on the bond, because Scotty had not committed a breach of contract, but that if the court held otherwise, Scotty was liable to Fidelity under the indemnity agreement. The towns counterclaimed against Fidelity for payment of the bond and cross-claimed for breach of contract against Scotty and the subcontractors. Other cross-claims were also filed but are not before us. The district court, on motions for summary judgment, held that Scotty had not broken its contract with the towns, and the court entered judgment for Fidelity on its main claim and the towns’ counterclaim and for Scotty and the subcontractors on the towns’ crossclaims. Although these orders did not dispose of the entire litigation, the district court certified them for immediate appeal under Rule 54(b) of the Federal Rules of Civil Procedure, so we have appellate jurisdiction. There are other parties below but they are irrelevant to this appeal.
In the normal course this lawsuit would have been kicked off by the towns’ suing Scotty for breach of contract and Fidelity for breach of its obligations under the bond. Such a suit would have had to be brought in state court with no possibility of removal to federal court, since the plaintiffs and one of the defendants, Scotty, would have been residents of the same state. Instead Fidelity precipitated the towns’ suit by bringing a declaratory judgment action against them and Scotty, an action in which the plaintiff was a resident of Maryland and all the defendants were Wisconsin residents. Fidelity’s invocation of the Declaratory Judgment Act, 28 U.S.C. § 2201, was wholly proper. The indemnity agreement gave it a potential claim of some magnitude against Scotty, but a claim on which Fidelity could not realize unless and until it was found to have defaulted on its obligations under the performance bond. Fidelity may have been concerned lest passage of time prevent its recouping from Scotty any money it might have to pay the towns. Only by forcing the towns to bring their action on the bond could Fidelity crystallize its own rights under the indemnity agreement. This is the kind of interest that the Declaratory Judgment Act was intended to protect. See, e.g., Illinois ex rel. Barra v. Archer Daniels Midland Co., 704 F.2d 935, 939-40 (7th Cir.1983). Joining Scotty as a defendant obviated any need to bring a separate suit to enforce the indemnity agreement in the event that Fidelity lost its suit against the towns.
*1266So far there is nothing to suggest that Fidelity and Seotty were not genuine adversaries. But Scotty’s answer, filed eight months after Fidelity’s complaint, admitted the principal allegations of the complaint and appears, though with something less than 100 percent clarity, to concede that if Scotty is found to have broken its contract with the towns it will be liable to Fidelity for any money that Fidelity has to pay on the performance bond though not necessarily for Fidelity’s attorney’s fees and other expenses of its litigation with the towns, which Fidelity’s complaint against Scotty also claims. But City of Indianapolis v. Chase National Bank, 314 U.S. 63, 73 n. 3, 62 S.Ct. 15,18 n. 3, 86 L.Ed. 47 (1941), holds that a dispute over costs and attorney’s fees is too flimsy a basis for preventing a realignment that will defeat the assertion of diversity jurisdiction.
Scotty’s answer denies that it broke its contract with the towns, but this is not a defense against Fidelity, which also denies Scotty’s breach, but against the towns. The potential conflict between Fidelity and Scotty over the latter’s contingent liability under the indemnity agreement was eliminated (except for costs and attorney’s fees) when the answer conceded liability, leaving Scotty with the defense that it had not broken the contract that Fidelity had insured. By staking its all on this defense, Scotty (of Wisconsin) aligned itself with Fidelity (of Maryland) on one side of the lawsuit against the towns on the other, thus putting Wisconsin residents on both sides.
Jurisdiction, however, depends on the facts as they exist when the complaint is filed rather than when the answer is filed, which in this case was months later. See, with specific reference to realignment, American Motorists Insurance Co. v. Trane Co., 657 F.2d 146, 151 n. 3 (7th Cir.1981); 3A Moore’s Federal Practice If 19.03[1], at pp. 19-52 to 19-53 (2d ed. 1982). Otherwise a party might take steps to defeat jurisdiction when he saw the case going against him — in a diversity case might, for example, move to his opponent’s state. So if Fidelity and Scotty were adverse parties when the complaint was filed, it would be irrelevant that something happened later to put them on the same side of the lawsuit.
For a similar reason it is irrelevant that the towns’ cross-claim against Scotty put it in the same boat with Fidelity. In this respect a cross-claim under Rule 13(g) of the Federal Rules of Civil Procedure is similar to a third-party complaint under Rule 14(a); and it is clear that if a case is properly within the diversity jurisdiction and the defendant files a third-party complaint against a resident of the plaintiff’s state the court does not lose jurisdiction over the plaintiff’s claim. Fawvor v. Texaco, Inc., 546 F.2d 636, 638 (5th Cir.1977); 6 Wright & Miller, Federal Practice and Procedure § 1444, at pp. 223-25 (1971). Although not consistent with the principle of complete diversity, this rule is required by the competing principle that jurisdiction should depend on the facts when the complaint was filed. The defendant is not permitted to defeat the plaintiff’s right to a federal forum by impleading a resident of the plaintiff’s state any more than he would be permitted to do so by becoming a resident of that state. True, if having sued a nonresident under the diversity jurisdiction a plaintiff later brings in a resident by filing a third-party complaint, the entire suit will be dismissed for lack of complete diversity. But that is because a plaintiff is not allowed to do in two steps what under Strawbridge would lead to dismissal if he did it in one by naming both defendants in his original complaint. Owen Equipment & Erection Co. v. Kroger, 437 U.S. 365, 374-75, 98 S.Ct. 2396, 2403, 57 L.Ed.2d 274 (1978). That is not the situation here, if there was enough “adverseness” when the complaint was filed to allow Fidelity to name Scotty as a defendant.
Although it is possible that even then there was “no actual, substantial conflict between the parties that would justify placing them on opposite sides of the lawsuit,” American Motorists Insurance Co. v. Trane Co., supra, 675 F.2d at 151, the mere possibility does not defeat jurisdiction. This would be clear enough if it were cer*1267tain that Fidelity, the Maryland resident, did not know when it filed its complaint that Scotty, the Wisconsin resident, would in effect confess liability under the indemnity agreement. To hold that federal jurisdiction was only tentative until Scotty filed its answer would violate the principle that jurisdiction depends on the facts at the time of the complaint. Although we cannot be certain that Fidelity did not know that Scotty would answer the complaint in the way it did, Fidelity would be unlikely to sue Scotty knowing that Scotty would confess liability — you do not ordinarily sue people with whom you have no quarrel — unless this was a means of collusively invoking the diversity jurisdiction. That is also unlikely. If Fidelity had reached an accord with Scotty and wanted to minimize its chances of getting booted out of federal court for want of complete diversity, either it would not have named Scotty as a party at all, and simply hoped that Scotty would not be deemed an indispensable party, which could destroy complete diversity, see Acton Co. v. Bachman Foods, Inc., 668 F.2d 76, 80-81 (1st Cir.1982) (more on this point shortly), or it would have asked Scotty not to concede in its answer that it was liable under the indemnity agreement. Fraud is too remote a possibility here to warrant going outside the complaint — which pleaded a genuine controversy between Fidelity and Scotty over Scotty’s liability on the indemnity agreement — in order to decide whether the district court had jurisdiction five years ago when this suit began.
City of Indianapolis v. Chase National Bank, 314 U.S. 63, 62 S.Ct. 15, 86 L.Ed. 47 (1941), the leading case on realigning parties to defeat diversity, is distinguishable. There (to simplify the facts slightly) a bank, acting as the trustee of bondholders of a gas company, sued the company and a third party, citizens of the same state, to enforce a lease between them. The bank and the gas company — that is, the bondholders and stockholders — had the same interest in enforcing the lease, which was advantageous to the company; and the Supreme Court, relying in part on the defendant’s answer, held that they belonged on the same side of the lawsuit. Although the bank had asked for a declaration that the lease was an asset from which the bondholders’ claims could be satisfied, the third party — the “real” defendant — had denied that there was any actual controversy between the bank and the gas company, and the Supreme Court agreed.. The bank “did not bring this suit in order to obtain a declaration that, regardless of the validity of the lease, [the gas company] is still ultimately responsible for the interest payments on its bonded indebtedness.” Id. at 73 n. 3, 62 S.Ct. at 18 n. 3. The naming of the gas company as a defendant rather than a coplaintiff was therefore “window dressing designed to satisfy the requirements of diversity jurisdiction.” Id. at 72, 62 S.Ct. at 18. There was a vigorous dissent which cast the facts in a quite different light, but we must be guided by the majority’s characterization and it is not applicable to this case. Scotty’s answer does not establish that the naming of it as a defendant in Fidelity’s complaint filed months earlier was “window dressing,” or that Fidelity would not have sued Scotty for a declaration of Scotty’s liability on the indemnity agreement except as a method — a curious method — of invoking diversity jurisdiction. So far as appears, Fidelity really did want a judgment declaring Scotty liable to it under the indemnity agreement.
There is, admittedly, some basis in the majority opinion in Indianapolis for requiring realignment whenever the plaintiff is allied on the primary issue in suit with the defendant to be realigned, though in conflict on a secondary issue. See 314 U.S. at 72, 62 S.Ct. at 18. By this test, the dispute between Fidelity and Scotty was secondary to their alliance on the question whether Fidelity had broken its contract with the towns. But such a test has not been applied in subsequent cases (our circuit’s decision in Trane, for example, requires an “actual, substantial conflict” but not that it be over the primary issue in the litigation), and we have recently expressed our distaste for spongy jurisdictional tests. See Illinois Dept, of Public Aid v. Schweiker, 707 F.2d *1268273, 278-79 (7th Cir.1983). The “primary-secondary” test would involve potential judicial diseconomy as well as uncertainty. If forced to litigate the issue of Scotty’s breach of contract in state court, Fidelity could still have brought a separate federal diversity action against Scotty for indemnity if it lost on the issue of breach.
It is not as if realignment had any systematic tendency to confine the diversity jurisdiction to the limits indicated by its policy of protecting nonresidents against local prejudice, on which see Chief Justice Marshall’s classic statement in Bank of the United States v. Deveaux, 9 U.S. (5 Cranch) 61, 87, 3 L.Ed. 38 (1809). If a state resident sues two nonresidents in state court and the nonresidents are prevented from removing to federal court under 28 U.S.C. § 1441(b) because one of them is realigned with the plaintiff, then you end up with a resident and a nonresident confronting a nonresident in state court; and the nonresident defendant will get no protection against bias in that court in favor of local residents merely because another nonresident is on the plaintiff’s side of the case. Cf. Currie, Federal Jurisdiction 131-32 (2d ed. 1981).
A ruling that Scotty’s presence in the case destroyed complete diversity would be peculiarly futile if Fidelity could have maintained a diversity suit against the towns alone, as it could have unless Scotty would be an indispensable party in any such action. See Fed.R.Civ.P. 19. If it would not be, and we order this case dismissed, Fidelity will be able to get right back into federal court by filing two separate diversity suits, one against the towns and one against Scotty and the subcontractors.
The definition of indispensable party in Rule 19(b) is restrictive, the required analysis highly particularistic. See Note of Advisory Committee on the Amended Rule, subdivision b; compare Acton Co. v. Bachman Foods, Inc., supra, 668 F.2d at 80-81, and Rapoport v. Banco Mexicano Somex, S.A., 668 F.2d 667, 669 (2d Cir.1982) (per curiam), with Greenspun v. Del E. Webb Corp., 634 F.2d 1204, 1210-11 (9th Cir.1980). Scotty has of course a big stake in a lawsuit in which its breach of contract, for which it may be liable to both the plaintiff and the defendants, is the key issue. On the other hand, so long as Fidelity vigorously litigates its position that there was no breach, Scotty is protected. And there is an argument that the intent of the parties to the performance bond was to allow suits by or against the surety to go forward unencumbered by having to bring in the contractor. So it is not certain that Scotty would be an indispensable party, and that is another reason for doubting that realignment is a particularly useful device for limiting diversity jurisdiction in accordance with its policy.
We note finally that even if Scotty would be an indispensable party in any suit between Fidelity and the towns, it does not follow that it would be an indispensable plaintiff, so that complete diversity would be destroyed whether or not it was actually joined. If it is indispensable, the suit cannot go forward without it, see Rule 19(b); but once it is joined, if it is a plaintiff in part and a defendant in part, realignment would not be required, under the view we take, unless the conflict with the original plaintiff was not actual and substantial.
So we have jurisdiction, and come at last to the merits, where the only issue is whether the district court was right in holding on summary judgment that the scrubber’s failure to perform as expected was not a breach of contract by Scotty.
By 1969 it had become clear to the towns that they did not have adequate methods of garbage disposal, and in 1970 they hired Donahue & Associates, an engineering firm, to study the problem. Donahue formulated a proposal, but before it could be approved the towns became interested in a different type of incinerator, one designed by a Dr. von Berlichingen. At their request he prepared plans for an incinerator that would have a “Venturi” scrubber. Von Berlichingen estimated the total cost of the incinerator, including scrubber, at $384,000. Bids were invited but when the lowest bid came in at $750,000 all were rejected. The towns then cast about for other designs but eventually decided that von Berlichingen’s was *1269the best after all. But they did not think he was competent to supervise the entire project by himself, so the contract they signed with him in 1973 made him hire Donahue as the engineer for the project. They also told him and Donahue to specify in the contract for the project a Krebs Elbair scrubber, or an “approved equal,” rather than the Venturi. The Elbair had never been used in a comparable project, but was cheap. The idea of using the Elbair apparently originated with Midwesco and Krebs Engineers (which is either the Elbair’s manufacturer or an affiliate of the manufacturer), both of which later became subcontractors for the project.
Von Berlichingen and Donahue drafted a contract and bids were again let. Scotty was the low bidder, was awarded the contract in 1973, and hired Midwesco to install the Elbair; Midwesco in turn hired Krebs to supply it. The incinerator was put into operation, Elbair and all, late in 1974 or early in 1975, but it repeatedly flunked emission tests given by the Wisconsin Department of Natural Resources and was shut down for good in 1977. The towns argue that Scotty promised that the scrubber would scrub well enough to enable the incinerator to pass its emission tests. Scotty argues that all it promised was that the scrubber would not be defective, and it is not. It operates in accordance with the manufacturer’s specifications; it just does not do the job the parties to the contract thought it would do. And there is no contention that the Elbair was not up to snuff because of something that Scotty or its subcontractors did or failed to do in installing it; it is just the wrong device for the task.
Scotty bid on a contract that specified the Krebs Elbair as the pollution control device. “The basic type of air pollution control equipment shall be Elbair, as manufactured by Krebs Engineers, or approved equal.... The scrubber shall be a single stage Elbair high energy wet gas scrubber Model PXH 10.10.2.” The expression “or approved equal” only appears to have given Scotty some leeway in the choice of devices, for the contract elsewhere states that the expression “gives the Contractor [i.e., Scotty] no authority whatsoever to choose or use any but the specified material. Where, for some good reason, a change in material might be required or desirable, the owner [the towns] may approve a substitution. Written approval from the Engineer [Donahue] is, however, required for such an effective change.” And where the buyer directs the contractor to perform the contract in a particular way and the contractor does so but disaster ensues because the directions given were wrong, normally the contractor is not liable.
The leading case in Wisconsin that so holds is Bentley v. State, 73 Wis. 416, 41 N.W. 338 (1889). (Fritz v. Woldenberg, 199 Wis. 99, 225 N.W. 700 (1929), is a similar case.) Bentley was hired to add wings to the state capítol under the direction of the state’s architect. He followed the architect’s plans but they were no good and the wings collapsed. The court held that the contractor had not broken his contract, because the state had implicitly promised that its architect would provide competent direction to the contractor. In other words, the contractor had not guaranteed a result over which he had no control.
People are reluctant to guarantee things over which they have no control — to assume risk, in other words (unless, of course, like Fidelity, they happen to be in the insurance business) — and this observation warrants a presumption that can be used to decide cases like Bentley where the contract contains no indication that a guarantee was intended. But such a presumption cannot be absolute. An important function of contracts is to allocate risk; and the performing party to a contract usually guarantees performance against at least some contingencies that are beyond his control. See, e.g., Field Container Corp. v. ICC, 712 F.2d 250, 257 (7th Cir.1983); Farnsworth, Contracts 538, 684-85 (1982); Holmes, The Common Law 300 (1881). If Bentley had been unable to complete the wings of the state capitol because of a strike, he probably *1270would not have been excused from his contractual undertaking; he would have been held to have assumed the risk of such an event when he signed the contract. True, a strike of his own workers would have been something more within Bentley’s control, or at least knowledge, than the negligence of an architect appointed by the buyer. That is why it was less likely that he intended to guarantee the buyer against the architect’s negligence than against the consequences of a strike. But if he had suggested the architect to the state in the first place, it would not have been so unlikely.
Even without going outside the contract — an excursion that would raise problems under the parol evidence rule if the contract was unambiguous and apparently complete — we can find intimations that Scotty guaranteed the performance of the Elbair. The section of the contract captioned “AIR POLLUTION CONTROL WORK” includes, under the heading “SCOPE AND PERFORMANCE,” the statement, “The Work covered by this section consists in furnishing and installing all work including the Air Pollution Control Work and in performing all operations in connection with the installation of an efficient State approved Air Pollution Control System.” State approval was what the system failed to obtain. The section continues: “The Air Pollution Control Equipment and process to be guaranteed to restrict dust emission to less than 0.05 grains .... ” This is the standard the Elbair later flunked. Under the heading “EQUIPMENT,” we read: “It is the intent of this specification to allow the Contractor to submit a bid on the equipment proposed to be furnished which will be a complete operating system and provide for stack emission which will meet the emission standards set forth in the previous section” (quoted above). Conceivably, all this language is just an expression of the towns’ hopes for the Elbair, but it can also be read as exacting a commitment from the contractor to install a system that would meet state emission standards, failing which the system would be completely worthless to the towns buying it.
The air pollution control section of the contract also contains a guarantee clause; and while for the most part it reads like a standard contractor’s guarantee against defects in materials or workmanship, it does purport to guarantee “performance,” and describes itself as “a performance guarantee based on operation.” Although the context and especially the limitation of the guarantee to five years suggests that compliance with the state emission standards may not have been the kind of performance the draftsmen of the clause had in mind, a subsequent clause states, “The Air Pollution Control assembly will only be accepted after the acceptable performance of same as indicated thru approved tests and after receipt of approval from the State Department of Natural Resources. Final payment for said assembly will also be subject to acceptable performance and such State approval of Unit.” (Emphasis in original.) The towns withheld final payment pending the approval that never came from the Department of Natural Resources.
It may seem fantastic that Scotty would guarantee the Elbair’s performance if the Elbair was foisted on it by the towns. Scotty is a construction company, not an environmental engineer. Why would it assume the risk that the Elbair would not work? But the towns spin a tale that if true would go far toward answering this question. They point out that the town fathers knew no more about air pollution control technology than Scotty and that von Berlichingen and Donahue were not experts in it either. They argue, with supporting documentation, that Krebs and Midwesco “sold” the towns on specifying the Elbair in the contract. Scotty replies that this is impossible — that the law forbade the towns to negotiate with anyone before putting out the contract to bids. But unlawfulness and impossibility are not synonyms. The towns’ allegations raise genuine issues of fact, which if material barred resolution of the controversy on summary judgment. See Fed.R.Civ.P. 56(c). The towns argue further that precisely because they were afraid to assume the risk that the Elbair would not work, they assigned the risk to the *1271contractor, who could reassign it to his subcontractors — to Krebs and Midwesco, who respectively supplied and installed the Elbair. The appeal record does not contain the contracts between Scotty and the two subcontractors, but there may be express or implied warranties running from the Elbair’s manufacturer by which the risk of the Elbair’s failure was shifted to the manufacturer, who unlike Scotty could not plead ignorance of air pollution control technology; and in a cross-claim not before us, Scotty is claiming indemnity from Midwesco.
So depending on what the facts really are, it is not so fantastic to imagine Scotty’s having assumed the risk of the Elbair’s failing as the price for getting the contract, maybe knowing it could shift the risk to the Elbair’s manufacturer — clearly a superior risk-bearer to the towns, which under the district court’s decision have been left to bear the entire risk. And if Scotty assumed the risk that the Elbair would fail, it is liable for the failure even though the contract did not allow it to use another type of scrubber. For a similar though not identical case where liability was found (we have, not surprisingly, found no identical case) see Mayville-Portland School Dist. No. 10 v. C.L. Linfoot Co., 261 N.W.2d 907 (N.D. 1978).
But we must consider whether the parol evidence rule bars a factual inquiry. In arguing that it does not, the towns rely heavily on Stevens Construction Corp. v. Carolina Corp., 63 Wis.2d 342, 217 N.W.2d 291 (1974). The contractor in that case, Stevens, undertook to furnish the concrete skeleton for an apartment house in accordance with the specifications of the owners’ architect. The skeleton was designed defectively and buckled after it was installed. It turned out that the architect had had no experience in the design of concrete systems and that Stevens, at the owner’s request, had retained a design firm to design the system for the apartment house. Interestingly, Stevens had extracted a promise of indemnity from its design subcontractor, as Scotty apparently did with its subcontractor Midwesco. Stevens argued that the fact that its subcontractor and not the owner’s architect had designed the concrete system that failed was inadmissible to vary the terms of the contract, which, Stevens contended, was unambiguous in that it included no guarantee of performance. But the Wisconsin Supreme Court held that the contract had a latent ambiguity. “The contract calls for Stevens to ‘furnish and install all ... precast basement beams.’ While such specifications appear clear on their face, when applying such language to the subject at hand, a patently ridiculous situation results — i.e., the contractor is to furnish undesigned precast beams knowing the contractor’s architect had no designing experience. Obviously, a literal reading of the contract which Stevens desires would lead to such an unreasonable result to suggest that the parties did not so intend.” 63 Wis.2d at 355, 217 N.W.2d at 298.
The concept of latent ambiguity may seem to do away with the parol evidence rule — which is no doubt why the towns rely on Stevens so heavily. If you may resort to extrinsic facts to determine whether a contract is ambiguous and the parol evidence rule therefore inapplicable, then the barriers to bringing in such facts are down from the beginning; on this view the parol evidence rule is applicable only if the inquiry into extrinsic facts yields nothing interesting. It is unlikely, though, that this is what the court meant in Stevens. The contract was not ambiguous in the sense of internally inconsistent, but it was silent on who had the design responsibility for the concrete system. The contractor might have had it or the owners’ architect might have, and which it was made all the difference to applying the principle of the Bentley case. The parol evidence rule applies only when the written contract is the complete and final expression of the parties’ agreement. Farnsworth, supra, at 451. If as in Stevens the contract is incomplete, the rule does not apply.
The situation in this case is different; it is that of a complete but equivocal contract. In specifying the air pollution control equipment to be used, the contract resembles Bentley, but on the other hand it contains repeated intimations that the contractor is to guarantee that the equipment *1272that the contract specifies will do the job, even though he had no say — no direct say, anyway — in the choice of equipment. These two sets of provisions are not necessarily inconsistent, since a contract party may, as we have said, guarantee results beyond his control, but their juxtaposition in this contract is an uneasy one, which makes it unclear how far the parties meant to go in extracting a performance guarantee from Scotty. Thus, the contract is ambiguous and parol evidence admissible. See Georgiades v. Glickman, 272 Wis. 257, 264-65, 75 N.W.2d 573, 577 (1955). The towns have made factual allegations that if proved would support, though not necessarily compel, see, e.g., Wood-Hopkins Contracting Co. v. Masonry Contractors, Inc., 235 So.2d 548 (Fla.App.1970), an inference that Scotty guaranteed that the incinerator would pass muster with the Wisconsin Department of Natural Resources. The allegations may or may not be true but the towns are entitled to an opportunity to prove them. The judgment of the district court must therefore be reversed and the case remanded for trial.