You're asking if the answer could also be not X - >B.
We have
P: A-most-B
P: ?
A-most-not X
We need to connect B with not X in a way to arrive at the conclusion
If all Bs are not X (let's say 100/100), and if most As are Bs (let's say 51 out of 100), then most As by definition must not be X. This works. B - >not X is our solution.
If we used not X - ->B instead, we can't conclude A-most-not X
Let's say 51/100 As are Bs (A-most-B)
Then we have all not Xs are Bs (100 out of 100).
What if the entire subset of Bs is a million? We could theoretically have no overlap with As and not Xs, which is why not X - >B does not work and doesn't help us arrive at the conclusion we're supposed to arrive at.
