The margin for conjecture 143 grows at approximately 0.25 per vertex (margin/n → ~0.25), reaching +152 at n=600. This linear growth distinguishes #45 from borderline counterexamples where the margin shrinks with n. The kite family exploits a structural gap between dense core eigenvalues and sparse tail properties — a gap that widens, not narrows, as graphs grow. This suggests a fundamental rather than edge-case violation.