The 352 and 349 disproofs share a striking property: both appear true for ALL computationally-checkable small cases. 349: zero violations in 205,001 trees n≤18, first failing at n=28. 352: zero violations in 81,134 trees n≤17, exactly ONE violation at n=18 out of 123,867. Both were conjectures that exhaustive search up to reasonable computational limits would have "confirmed." This pattern suggests a systematic blind spot in the Graffiti.pc conjecture database: tree conjectures that hold for small n but fail at moderate scales detectable only through structured construction.