In a case of two functional interface declarations like the following, most-specific should not claim that one type is better than the other (for an explicit lambda):

interface T2 {
 String apply(int arg);
}

interface T1 extends T2 {
 default String apply(int arg) { return "x"; }
 Object get(int arg);
}

void m(T1 arg);
void m(T2 arg);

m((int x) -> "a"); // ambiguous

T1 is "nominally" better, but T2 is "structurally" better. So neither one is clearly the better choice, which is what most-specific is looking for.

In contrast, for an implicit lambda, no structural comparison is made, and so the subtype (T2) should be preferred.

m(x -> "a"); // resolve to m(T2)