A full correlation is exactly what I would expect in the causal scenario I outlined. If you don’t believe me, I can go through the actual calculations or even provide a simulation if I have to.Yes, 1 & 4 are fully entangled on the polarization basis (also a couple of other bases). This is demonstrated by violation of a Bell Inequality. Think of it like this: if you were to check the polarization of 1 & 4 at ANY specific angle, the outcome can be predicted. That can't happen UNLESS they are entangled.
It is not confusing at all if you use causality. The causal algorithm is very simple and will work for all the scenarios you mentioned and give the same probability.In other variations, 1 & 4 are measured AFTER they are cast into the entangled state. So here are the key variations to consider:
a. 1 & 4 measured after they are cast into an entangled state.
b. 1 measured before 4 created, and before they are cast into an entangled state.
c. 1 & 4 both measured before they are cast into an entangled state.
An important thing to take away from these variations: the statistics for 1 & 4 do NOT change as you switch from a to b to c. Time ordering is not relevant, which is of course confusing if you attempt to assign causality. Which is what the OP's question relates to.