Dear Joy,
I think I figured out what is going on. The EPR completness criterion simply means that one gets the entire state space, be it phase space, Hilbert space, or other spaces. By considering actuals al well as counterfactulas you are then in fact ignoring the time evolution and get the entire state space. In general, any state space will obey EPR's completness criterion. (Try your approach on classical mechanics for a hypothetical hidden variable theory.)
Local realism in your approach is only factorizability. Factorizability is the opposite of entangelment and this gives the approach its classical intuition. Factorizability is not local realism as locality was considered by Einstein and Bell to mean just that: spatial separation. So is Bell's theorem invalid? No, because a Hilbert space dimension is N^2 for N psi(s) and it is not always separable. Still, you manage to arrive at separability. But this is not done directly in the original Hilbert space. For example you need to embed Bloch sphere in S3. S3 becomes then a different kind of state space and in fact you are rewriting QM in a diferent state space with a different formalism. But wait a minute. Is QM not supposed to be uniquely written in Hilbert space formalism? How about Piron's result of recovering Hilbert space over division algebras from propositional logic? The answer is no as there is a counterexample to that: QM in phase space via Moyal bracket.
So your prescription for separability is embedding (if possible) the Hilbert space inside S0, S1, S3, or S7 to achieve parallelizability. Because this contradicts the nonseparability of the standard Hilbert space, this means that QM over S0..S7 is something qualitatively different than a standard Hilbert space formulation. (And indeed, in your formalism you use different mathematical objects.)
Also it is not clear if this formulation of QM goes beyond QM or not. In other words, can you always succeed in embedding any Hilbert space in S0..S7? Probably not based on dimensional analasys for higher dimensions.
Another issue. If Bloch sphere is embedded in S3, would not this mean that we still deal with traditional complex QM? Let's look at another example first. Real quantum mechanics can be embeded in complex QM, but the meaning of the wavefunctions is qualitatively different. Probably something along similar lines is happenning here, I don't know. To get a better grip, an analasys of time evolution might clarify things as time tends to dissapear from the picture as both actuals and counterfactuals are considered. Maybe this analasys will show that you are still in a traditional Hilbert space (the spin factor case), and that the Killing flow is what you traditionally obtain in the original embedded space.
If I were to venture a guess, in S3 the time flow is not the same as in Bloch sphere and the meaning of psi does not stay the same. For if they do stay the same, all possible time evolution in Bloch sphere would be enough to achieve EPR completness which is not the same as S3 is needed.