Richard,
The specific slip Judy identified you'd made was in misunderstanding exactly what 'quality' weak measurement (WM) statistical correlation was quantifying. It seems a more common misunderstanding that I'd realised. Before clarifying that I also clarify that I do well understand Bell's tautological point and don't dispute it. What I identify is the flawed assumption of QM on which it's founded. Correct that assumption, so allow B to find a OR b whatever A finds, and Bells written statement is circumvented.
WM statistics fool many. What they do is compare what distribution A finds at any one angle with what B finds at that angle, giving an output only of what we might call 'sameness' of finding. I derive that quite separately from the initial 'local' (single detector) inequalities in may paper for that reason.
A few worked examples should help comprehension.(You should note we're now well beyond the H/T toss as it's the same in either case of H/T vs A/B).
Let's take all cases of B where A chooses 0^o and finds GREEN with 100% certainty.
B chooses 180 so finds RED with 100% certainty. WM finds 0% A/B correlation.
B chooses 0^o so finds GREEN with 100% certainty. WM finds 100% correlation.
B chooses 90 so finds SAND, so only 50% certainty, so WM finds 50% correlation.
B chooses 22.5^o, = LIME so ~95% decide GREEN, so WM finds 95% correlation.
B chooses 45^o, = YELLOW. Trickier, but ~77% say Green, WM = 77% correlation.
You'll find the same result for ANY choice by A. "Correlation" is then only a measure of 'sameness' of A and B findings. The original inequality is inherent in each original set of findings, as single detector experiments show. The truth is then quite different to the claims made about what statistical analysis produces.
The simple classical experiment then reproduces the quantum prediction because 'flipping' field electrons also flips the detected spin direction.
Peter
PS; I've update the experiment protocol (EPS not Randi) kit with instructions, attached below, but see the paper appendix A. I probably also need to reset the cosines at your 90^o datum to make the simple mathematical derivation clearer. Do run through the experiment and give me your view.