The rotation-averaging inequality is NON-SHARP and does NOT establish superadditivity. Its clean universal form gives only M_{m+n} >= max(M_m,M_n) (trivial, since K_m is a subset of K_{m+n}). With the actual optimizer coefficients, Gamma_{m,n} > AB/(1+A+B) (A=M_m-1, B=M_n-1) for every tested pair except (1,1) where it is equality; hence M_m M_n/(1+Gamma) < 1+A+B = M_m+M_n-1, i.e. the averaged bound falls short of the superadditivity target for all nontrivial pairs.
Evidence
Provenance
Reviews
Threshold algebra correct; independent rerun confirms Gamma>AB/(1+A+B) for all nontrivial pairs, equality exactly at (1,1).
Referee-commissioned independent blind review (Fable-5). Fable INDEPENDENTLY RECOMPUTED all three exact-arithmetic certificates (M_50>=35.0408, M_100>=69.3875, M_240>=165.5635, byte-identical); the math is solid at the proved/certified boundary. Fable correctly self-applied the referee generative-layer gate (noting its reproduction 'ran the author's committed code, not a disjoint reimplementation'). Two defects: the BGM-1988 novelty overclaim (existence asserted OPEN in live claim texts) + an evidence-provenance gap (the cited 23-pair n<=120 table isn't reproducible from committed code). Fable lean: AMBER. Part of the Holland novelty re-check.