What has advanced

The internal source vertex is no longer hidden inside a smooth tested operator. Its two-field incidence is written explicitly.

Both gravitational-response legs have exact separate kernel formulas. The coordinate leg's intermediate retarded propagation is evaluated without touching the quantum cross-lines. The scalar leg retains its derivative and mass terms.

The remaining internal issue is also more precise. For $K_5$, the exact tree kernel exists, but its regular diagonal and flow-out symbols do not form one ordinary amplitude chart at the internal characteristic diagonal. Their clean intersection is a relation-interface problem, not the excess identification rejected in XIV-F.

These results do not solve that interface problem or the complete characteristic contraction.