Dear Edwin,
To say that it is "de facto true" that Bell's example about spin is just illustrative and no part of the theorem does not address the point. The entire discussion of the detailed model makes no contact with the theorem. The theorem holds of any theory at all that is local (in the sense Bell articulates) and makes certain statistical predictions. That these are predictions about anything called "spin" or anything treated quantum-mechanically is no part of the theorem at all. All one needs are the conditional probabilities for outcomes of certain experiments, which need not be described in any more detail than "Instrument 1 is set to setting A" and "the outcome is outcome 1" or "outcome 2" One can use "spin measurements" in quantum theory as instances of this sort of thing, where the setting is the orientation of the Stern Gerlach magnet and the outcome is a spot on a screen appearing in one place or another. Clearly any theory at all might make such predictions. Since the theorem is only about these sorts of conditional probabilities, it is in no way "about" quantum theory.
Your initial characterization of the question Bell was asking is not accurate. He was not asking whether one could somehow find a theory that predicts the outcomes of experiments deterministically, he was interested rather in whether any local theory at all (deterministic or probabilistic) could recover a certain set of predictions. He insisted on this many times, and complained that his point had been almost universally missed. In fact, the paper relies on an understanding of the EPR argument, which had already established that locality can only be recovered in a situation with perfect EPR correlations if the theory is deterministic, but, as Bell says, "It is important to note that the limited degree to which determinism plays a role in the EPR argument, it is not assumed but inferred. What is held sacred is the principle of 'local causality'-or 'n o action at a distance'. Since the EPR correlation are recoverable by a local theory only if it is also deterministic, one can then ask about constraints on such theories. Bell demonstrates such constraints.
On p. 4, you list what you call "Bell's key physical assumptions". None of these are assumptions or premises of his theorem. The theorem applies to any situation in which the outcomes of certain experiments can be categorized as, e.g., "outcome 1" or "outcome 2", and correlations between the outcomes on different sides predicted. The theorem, which is not particularly about spin, has none of these assumptions as premises, so no discussion of them can have any significance for the theorem.
What is particularly odd about your presentation is that you claim that Bell has a "hidden constraint" in his proof, but nowhere actually discuss the proof itself, but rather only the illustrative example. It would help if you would actually point out where in the proof the supposed constraint appears. Your rather extensive discussion of the toy model makes no direct contact with the theorem itself.
As for your own model, let me try to understand the claim that you make. Your equation 4 has the consequence, as you say, that the deflections produced by Stern-Gerlach magnets will not be quantized, that is, that we cannot, as a practical matter, distinguish the outcomes into two classes, usually denominated "spin-up" and "spin down", determined by the location of the detected particle. If that is correct, then your model certainly does not reproduce the actual phenomenology reported in the lab, nor the predictions of quantum theory. Since the correlations discussed by Bell are correlations between the outcomes on the two sides, which are taken to always be either "spin-up" or "spin-down", and since these are also the predictions of quantum theory, then it would appear that your model actually makes no contact with Bell's topic. You do not explain how the top graph on p. 7 was created, or even what it means. Here is a key sentence from that page: "If I throw away this θ -information by truncating the measurement data, i.e., setting the results to A, B = ±1 , my constrained model cannot produce the correct correlations." The obvious reading of this sentence is that in your model, the outcomes of the experiments are not categorized into two classes, spin-up and spin-down outcomes, and that if one requires such a categorization of the outcomes then you get Bell's result. But it is an observed fact that the outcomes do sort into these two classes, and it is a prediction of the quantum theory that they will, and furthermore if they do not then it is not at all clear what the meaning of "correlation" in your theory is, since the predicted correlations are between these binary results. So you it would help if you could do these things:
1) point out where the supposed "hidden constraint" actually appears in Bell's theorem.
2) Explain, if your model does not predict quantized outcomes for spin experiments, what bearing it has on quantum theory, or Bell's theorem, and what you even mean by a "correlation" between the results on the two sides.
Regards,
Tim