You can not try different settings at A in a single given experiment. Whatever the setting at A, B will get 1 or -1 with probability 1/2.I see your point: in your case, if B would be -1, A cannot other than be -1 too, so the outcome of A would have been unchanged. And how can something that did not change have an effect, right? But I'm not speaking of outcomes of A, but of settings of A. Given a different setting at A, would that allow for a different outcome of B?
I suspect that the above sentence does not satisfy your concern, but I still don't know why.