UPDATE: I asked Victor Ostrik, and he suspects this argument is circular, in that some of the references I've cited use implicitly the commutativity of the diagram. Also, I managed to figure out a proof for the subregular case (that won't work in general).