It's because measurements at B do not convey any information to C while the measurements are performed and vice versa. Unless B calls C to inform them of the choice of measurement setting, C will not know the measurement outcome at B's side. This is true even if they know that they share entangled states prior to prior to performing measurements.
> Edit: Also, to add, my understanding is that they are not "getting angles" per se, but would be seeing distributions. Which is why you would need more than 1 particle, as it were. So, you would say of the X I have recorded, 30% have been blue, 70% have been green. I suppose the concern is that you have no way of knowing when the "100%" mark is done until after classical communication, such that it is impossible to know what the final distribution you are measuring is? Effectively?
There has to be post processing of data where they drop the results of rounds where their measurement choices don't match. This is important because of what's called non commuting measurements. Measurements in one setting don't give us any information about measurement outcome in another setting. So effectively at one end, they have to record their measurement choice and corresponding outcomes of said measurement. And when comparing the data, the participants only have to keep the outcomes of rounds where the measurement choice is the same at both end
Framing I saw was more of a truth table like where B/C have known states they can be in that each lead to a known distribution of outcomes. It was not clear that the known distribution was only an observed distribution.
I'm not sure I'm understanding your percentages statement completely, but when the parties have entangled states, these inconsistencies will match on each side under the assumption that their measurement choice is the same.
> Framing I saw was more of a truth table like where B/C have known states they can be in that each lead to a known distribution of outcomes. It was not clear that the known distribution was only an observed distribution.
If the states are locally known to the parties, they'd still have to perform a statistical run of experiments as before. The choice of measurements would be key in distinguishing a classical from a quantum correlation
(And just to rapidly get it out there, I fully expect that I'm merely misunderstanding something here.)
Edit: Also, to add, my understanding is that they are not "getting angles" per se, but would be seeing distributions. Which is why you would need more than 1 particle, as it were. So, you would say of the X I have recorded, 30% have been blue, 70% have been green. I suppose the concern is that you have no way of knowing when the "100%" mark is done until after classical communication, such that it is impossible to know what the final distribution you are measuring is? Effectively?