That theory is inconsistent. Can it make valid predictions before Bob measured anything? Are those predictions still valid if Bob measured something else? Then you can't avoid CFD. CFD is a logical necessity of any theory that makes valid predictions for experiments that might not be performed irrespective of ontology.The view according to retrocausality (experimental outcome is part of the computation as in path integral) is that you have some sort of field or graphical or ... structure in the spacetime region between Source and sink. This field/graph exists in spacetime (as opposed to configuration space), so it can have ontic status in the normal sense. The field/graph relates all elements of the experimental procedure and is unique for different detector settings, so CFD is meaningless even though there is a fact of the matter concerning the ontology in that region of spacetime. So, if Bob doesn't measure "a," then the field/graph structure in the spacetime region of the experimental procedure is different than if he had measured "a." In other words, you're working in a blockworld (spacetime) perspective with a particular outcome (again, as with path integral) rather than a dynamical, time-evolved perspective in configuration space (as with Schrodinger's equation).