Of course people might not necessarily act rationally but you've got to draw a line somewhere, otherwise elections are simply undefined behaviour and talking about properties of elections becomes meaningless.
But, no, you cannot safely recalculate that result, because by and large people are the representation of chaos. undefined behavior is the norm, not defined behavior. Defined behavior is only present in any of our societies because of 'social norms', which is the only reason I relent and say that some assumptions might be possible. But I will not agree that you can 'safely' recalculate that result just using whatever assumptions some person arbitrarily has deemed 'good enough to trust'.
That kind of thinking is why most polls are actually wrong, but taken as good enough. (Yes, I'm one of those people who nit picks the randomness of their selection biases.)
P.S.
All it takes to swing an election in favor of the underdog is a single new revelation. All it takes is for people to suddenly question their previous choices. A & B might be their norm; but people can be swayed to choose C or D if the situation is right.
This is why we can't really trust assumptions without some sort of evidence to back them. At which point they stop being assumptions...
Otherwise you're just arguing the vote is invalid because it would have ended differently in a rerun, and you have to stop somewhere.
In normal elections we see this with people who don't like any choice, but don't want to spoil their ballot. So they pick their next best favorite.
This means that if A is removed, people might pick B or C instead just because A isn't there.
People are tricky. Hard to pin down what's exactly going to happen every time. There are distinguishable patterns and such; but ultimately you can't rely on them as if they are infallible.