It's that "only" in "... only in the sense that ..." upon which the whole thing hangs. Presupposing unitarity (and locality) and constraining models to conform to that presupposition may be that complication, that blinder, that prevents discovery of the simpler underlying rule from which either or both emerge. And either or both may be necessary consequences of the much simpler, underlying rule (if there is one to be found), at least on the observable side of any predicted phenomena, but that can't be determined unless one is willing to first throw away the constraints that keep you from finding it. (In effect, it's analogous to simply adding two natural numbers directly rather than using Peano arithmetic and checking at each iteration that an increment occurs. Either way you will arrive at the same answer, but the unconstrained rules for addition allows for negative numbers and fractions.)