Too small a bound: an Evidence Press briefing. Every graph, a set of points joined by lines, has a complement: the same points, joined by exactly the missing lines. Each graph also has a list of numbers, its Laplacian eigenvalues, which measure how well connected it is at different scales. How large can the same-ranked eigenvalue of a graph and of its complement be together? In twenty twenty-five, Chen, Guo, Li and Wang conjectured a limit on their product: for n points, at rank k, no more than n times n minus k. They proved it for the top rank and for ranks from a half to three quarters of n. The lower ranks were open. This paper shows the limit is false. The trick is balance. Make half the points a clique, all joined to one another, which lifts the graph's eigenvalue. Leave the other half unjoined. In the complement they become a clique, which lifts the complement's eigenvalue. Then add a carefully balanced set of lines between the halves, so that both numbers stay high at the same rank. The example has fourteen points, arranged using the Fano plane, the smallest finite geometry. At rank six the product is about one hundred and fifteen point eight five. The conjecture allows only one hundred and twelve. The paper then finds the true largest product, to leading order, for almost every rank, with a sudden drop at the middle rank. Gibson had already shown that the conjecture's description of equality fails, and left the lower ranks open. A numerical search of every graph up to ten points found no violation, but only the bound of fourteen points is proved, and some ranks remain undecided. This is an anonymous, AI-assisted candidate. It has not been peer reviewed, independently reproduced or formally verified. Published on the fourth of November twenty twenty-six, its paper and evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.