This kind of situation isn't special. It comes up in every election, and approval is completely confused about how to handle it.
I recently had the occasion to really game out what happens in ranked choice voting. It is similar to approval voting (you should only rank people to approve of), but in practice you elect whoever gets past the 50% threshold first, which may not be whoever had the most raw approval. The first person past 50% is effectively random, especially in a large field with complicated voter preferences. So, you get an illusion of choice: everyone lists their preference, but the actual winner is a somewhat random choice amongst the candidates with >50% approval. Thus, I advocate removing the illusion, and simply picking the candidate with the greatest approval. Easy.
"What the voter should do to get the result they want" assumes a level of individual agency which really isn't present in a real election: The result is always a combination of individual choice with collective will.