I have read the papers, but I think what will be more helpful is for you to simply state a fixed, deterministic HV that is a counterexample. This disagreement is not really about the TI, it is a much more general question of what it means for anything to be a hidden variable.
My argument is very simple: in a deterministic hidden variables interpretation, you must assign hidden variable which uniquely predicts the outcome of each local measurement. Otherwise, local measurements cannot be deterministic. If local measurements are not deterministic, the interpretation is not deterministic. The Bell's theorem non-locality only enters as a means of correcting the HVs when necessary, reacting to other local measurements performed elsewhere. This will be a de facto pilot wave.
If, as you say, "the fixed deterministic value is not "A is up and B is down" or anything like that" then I do not see what it can even mean to call these values fixed or deterministic.
Arnold has suggested:
But I can just turn this around into: "the value at spacetime point x is C(x)" and "the value at spacetime point y is C(y)". As long as C(x,y) maps to a pair of unique values (not a statistical mixture at each local site), this is clearly isomorphic to the above. If C(x,y) maps to statistical mixtures at each local site, this bilocal variable is not a deterministic HV for local measurements.
Dar, you said:
If each individual observation of ⟨A⟩ and ⟨B⟩ reveals a preexisting hidden variable of ⟨A⟩ or ⟨B⟩, the joint stats will not violate Bell ineqs without assuming something more (pilot wave, superdeterminism).
If there are no preexisting hidden variables for ⟨A⟩ and ⟨B⟩ individually, local measurements are not deterministic, and this is not a deterministic HV interpretation at all.