This suit arises out of a struggle between Raymond Allen and Sanford Takiff for control of a national bank. It was filed by Allen, initially against the bank and its president, complaining that the bank’s action in adjourning the shareholders’ meeting at which the board of directors was to be elected in 1996, and in subsequently seating a board dominated by Takiff, violated provisions of the National Bank Act, 12 U.S.C. §§ 61, 71, regarding the orderly governance of national banks. Allen asked for an injunction commanding the bank to seat a board in accordance with the ballots that he had cast at the meeting. Takiff was allowed to intervene in the suit under Fed.R.Civ.P. 24(a) as a defendant and file a counterclaim against Allen under the ancillary jurisdiction of the federal courts (now a part of their supplemental jurisdiction, 28 U.S.C. § 1367(a)), claiming that Allen was entitled to no relief because he had cast votes for directors in such a way as to violate an agreement between the two men, and seeking an injunction commanding Allen to comply with the agreement as Takiff interpreted it. (The parties do not make clear what state’s law governs the interpretation of the agreement, but it appears to be Illinois law, and we shall so assume.) The district judge conducted a bench trial. Although testimony and other evidence going beyond the text of the agreement were introduced, the judge concluded that he didn’t have to resolve any evidentiary disputes because the agreement was unambiguous after all. He agreed with Takiffs interpretation and ordered Allen to cast votes in future elections for directors in accordance with that interpretation.
*286Allen and Takiff are equal owners of a holding company that owns the bulk of the shares in the bank. But Allen is the president of the company and under its bylaws votes the shares in elections for the bank’s directors. Takiff, and Takiff family trusts that he appears to control although he is not the trustee, own additional shares in the bank; we shall call these Takiffs “directly owned” shares. The bank’s board of directors has fourteen members, and Takiff claims that Allen agreed to cast an equal number of the holding company’s bank votes for seven board candidates nominated by Allen and seven nominated by Takiff. If Takiff is right, then by concentrating the votes of his directly owned shares on two additional candidates he can elect nine directors, leaving Allen with only five. Allen claims the right to concentrate the holding company’s votes on his seven nominees so that he can defeat the two additional Takiff candidates.
The holding company, P.N.B. Financial Corporation, had been formed by Allen and Takiff in 1978 to acquire the bank. Allen was made president of P.N.B. with power as we noted to vote its shares, and Takiff vice-president. P.N.B. owns 84 percent of the common stock of the bank, Takiff and his family 13 percent, Allen only .2 percent, and other shareholders the balance of less than 3 percent. Eventually the partners fell out, and in 1991 Takiff brought a suit in state court under section 12.50(b) of the Illinois Business Corporation Act, 805 ILCS 5/12.50(b), against Allen and the holding company to dissolve the holding company and distribute its shares in the bank to Allen and Takiff, which would give Takiff 55 percent of the bank’s common stock. The suit was settled the following year. The holding company was not dissolved, but the settlement agreement provides that “P.N.B. will nominate (to the extent not nominated by others), and all its shares in the Bank will be voted for, election of seven directors nominated by Allen ... and seven directors nominated by Takiff.”
The National Bank Act ordains cumulative voting for directors. 12 U.S.C. § 61. That is, each shareholder is entitled to cast votes equal to the number of his shares times the number of directors to be elected and he can distribute his votes among the candidates as he wishes — he can cast all his votes for one candidate, or spread them evenly over all his candidates, or do anything in between. If the holding company spread its votes evenly among the seven nominees of Allen and the seven of Takiff, each of the fourteen would get 86,000 votes. This would enable Takiff to knock off two of Allen’s nominees with two additional nominees of his own, because the bank shares directly owned by Takiff have 188,000 votes, enabling him to cast 94,000 votes for each of two nominees. Takiff saw this possibility and caused two additional candidates to be nominated in order to break the deadlock with Allen and seize control of the board. To defeat Takiffs maneuver, Allen cast 95,000 votes for each of his seven nominees and only 76,000 for each of Takiffs original seven and none for the additional nominees. By doing this, he defeated — or would have, had the meeting not been interrupted — not Takiffs additional nominees, who had more votes than any of Takiffs original ones, but two of the original ones, each of whom received only 78,000 votes (76,000 votes cast by the holding company at Allen’s direction and 2,000 from other shareholders). The district court held that Allen’s maneuver violated the settlement agreement.
The briefs range widely over the evidence introduced at trial — for remember that the district judge did conduct a trial, even though he concluded that the agreement was unambiguous and therefore the evidence that had been presented at trial was irrelevant. The district judge thought it clear that the agreement required Allen to distribute P.N.B.’s shares equally across seven Allen and seven Takiff nominees rather than to maneuver so as to prevent Takiff from using the bank shares that he owned directly to elect a majority of the board. The judge was probably right in his interpretation (the qualification is critical), if only because Allen’s interpretation makes the agreement ambiguous and thus invites litigation. Allen claims the right to decide, subject no doubt to judicial review, that candidates not nominated by Takiff are really Takiff nominees, entitling Allen to withhold crucial votes from the official Takiff nominees, that is, the nominees *287submitted by TaMff to Allen in the latter’s capacity as president of and vote-caster for the holding company. Suppose TaMff sold the shares he directly owns, as of course he might. The purchaser would have 188,000 votes to cast for his own candidates, and suppose he cast 94,000 for each of two. Under Allen’s view of the settlement agreement with TaMff, Allen could, if he deemed the purchaser an ally of TaMff, distribute the holding company’s votes in such a way that all Allen’s nominees were elected but that two of TaMffs official nominees, having only 76,000 votes, were defeated by the new shareholder — who might be (or become) not an ally of TaMffs but an ally of Allen’s, giving Allen a nine to five margin over TaMff in the bank’s board of directors. This prospect would make Takiff very reluctant to sell his directly owned shares; yet under Allen’s interpretation of the settlement agreement, these shares though controlled by Takiff give him no increment in voting power, for although TaMff (including his family trusts) owns directly or indirectly 55 percent of the bank and Allen only 42 percent, Allen claims the right to deadlock the board of directors. A court asked to evaluate the consistency of Allen’s action with the agreement would not be obligated to accept Allen’s view of who was an ally of whom — but how likely is it that Takiff agreed in the settlement agreement to make the identification of alliances a litigable issue? Not very, especially with the danger of corporate paralysis that is implicit in an agreement that creates a deadlocked board of directors.
But to be certain that Allen’s interpretation is wrong, we would have to know more about the likely outcome of the suit for dissolution. If Allen was likely to have won it, this would have given him the leverage to obtain the deadlock. For unless the holding company was dissolved, Allen would continue to enjoy the right to cast all the votes of the holding company in favor of the candidates of his choice, giving him control over the board of directors of the bank because the holding company owns 84 percent of the bank’s common stock. If instead the suit to dissolve the holding company and distribute its holdings of bank stock to Allen and Takiff was quite likely to succeed, giving Takiff control of the bank, then TaMff, whose lawyer drafted the settlement agreement, would surely not have agreed to give Allen the room to maneuver that he claims, had the issue arisen in the negotiations. The record contains no evidence concerning the likely outcome of the dissolution suit. But it is germane to note the uncontradicted evidence that had the holding company been dissolved, Allen and Takiff would each have faced a $2 million tax liability, and this is a reason to believe that Takiff brought the suit to induce a settlement with Allen rather than to dissolve the holding company and take over the bank; so this would be a bit of evidence in support of Allen’s interpretation.
All tMs argumentative to and froing would be of no moment if the agreement were crystal clear, Frydman v. Horn Eye Center, Ltd., 286 Ill.App.3d 853, 222 Ill.Dec. 151, 155, 676 N.E.2d 1355, 1359 (1997); McElroy v. B.F. Goodrich Co., 73 F.3d 722, 727 (7th Cir.1996) (Illinois law), or if the extrinsic evidence interpreted as favorably to Allen as the record permits showed that the judge would have committed a clear error had he accepted Allen’s interpretation. Taracorp, Inc. v. NL Industries, Inc., 73 F.3d 738, 743 (7th Cir.1996) (Illinois law); Smart v. Gillette Co. Long-Term Disability Plan, 70 F.3d 173, 178 (1st Cir.1995); E. Allan Farnsworth, Contracts § 7.14, pp. 538-40 (2d ed.1990). Neither condition is satisfied. The agreement does not say that Allen is to cast the same number of votes for each of the fourteen nominees, and, although the implication is that he is to do that, it may conceivably be overborne if the effect would be to defeat the even division of the board that the agreement seems also to contemplate. The evidence that might disambiguate the agreement is either disputed or simply missing. In retrospect, since the district judge conducted a trial it would have been a good idea, in order to avoid a succession of appeals, for him to have decided the case not only on the assumption that the agreement was clear on its face but also on the alternative assumption that it was not. He did not do so, so we must remand for further proceedings consistent with tMs opimon.
VACATED AND REMANDED.