Thanks Declan, your prompt reply is appreciated. It's also good to see that we have some agreements; but I won't dwell on them for now. Instead I want to discuss what looks like (in my opinion) a serious point of disagreement.
Please note that I have no wish to discourage you -- quite the contrary -- because I think you have guts and brains; and perhaps it is me that errs. However:
Imho, what you call "Classical" or "Classical Physics" is not classical at all.
Thus your "reason for the Classical prediction being the blue line is this: Classically each detector has a semicircle of directions where an incident photon will give a + result, and the other semi-circle (of the whole circle) where an incident photon will give a - result."
May I take it that this "classicality" is part of your own theory? Or do you have a source? And can you be more specific, please, and consider your "detector" to be built from a polarizer followed by an analyzer?
For it's true that Bell 1964:(4) uses a similar approach, but only by way of illustration: for I'm not aware of any classical textbook advancing such a theory. What's more I do not see how your idea works for the usual classical demonstrations that are conducted with three 'sandwiched' polarizers: where brightness measurements show good accord with classical theory without allowances for "non-detects"?
You should be able to do the classical textbook calculation and see that it yields an expectation of one-half the QM value; which is NOT the blue line: instead it will be one-half the green line.
Then, regarding this next point of yours [with my emphasis]:
"Essentially it seems to me [DT] that you [GW] are saying that the two photons in the experiment have opposite angular momenta, thus conserving angular momentum across the experiment. Yes, there is no doubt of that - but this is not sufficient to assure that detectors A and B have correlated results at different angles, as each detector has a probability of detecting each photon as either + or -. What the EPR experiment reveals is that when the two detectors have nearly the same orientation they have a high degree of correlation despite not knowing where the other detector is. So to build up a high correlation between A and B, each detector would have to register more + results (for photons incident on them from at the same angle) when the other detector is in a certain location; then register more '-' results when the other detector is in a different location, despite not being able to know that other detector's location!"
In reply, with Einstein-locality ensuring that no detector has any 'knowledge' about the other: in EPRB (eg, using Aspect's experiments) the probability of +1/-1 from each detector is 50/50, for all (a, b); so there is no "knowing" required. And the related correlation is twice the classical correlation because pairwise "entangled" photons (ie, in the singlet state) are more highly correlated than pairwise correlated photons (in beams) correlated by linear-polarization only.*
Re the latter, I recommend that you do the classical calculation; re the former I would encourage you to study my essay and ask questions. For I am keen to see where we might disagree and where things might be improved; me noting that the only change I make to modern physics is to take Bohr's "disturbance dictum" seriously.*
* My own dictum: Correlated tests on correlated things produce correlated results without mystery; and correlated tests on more correlated things produce more correlated results without mystery.
PS: The GHZ I mentioned is [14] in my References; you'll see the 4-particle GHSZ variant of EPRB in [13].
HTH, with best regards; Gordon