It's better to force the opponent to capture than give him the choice of capturing or doing something else. If your opponent chooses not to capture the rook it's because he has found a move he thinks is even better than taking the rook. And that move is then something you should fear.
A big part of chess is maximizing your own choices and freedom while restricting your opponents choices.
Level 1, yes, this is the correct way of thinking. You should always take the highest scoring move.
However, Level 2, making a decision harder for your opponent, might force them to spent more time thinking about the decision. If for some reason there is an imbalance of pondering, this might be beneficial. Suppose that you knew X position would be reached before your opponent, so you had more time to study it, you know what the correct piece to take is, whether a promoted rook, or a previously existing rook, but your opponent doesn't yet, and crowning to a queen will force your opponent into a move without a thought.
The computer will sometimes do this, more because of randomness than strategy, but it is probably always the case that if they underpromote, you should take, it's like a tell of theirs. Perhaps there is a case for nash equilibirum where you must sometimes offer an underpromotion in a scenario where a queen would have been marginally better such that underpromoting doesn't signal to your opponent that they should take the piece (whichever it may be, I'm a bit dizzy)
Very theoretical, but not impossible that underpromoting in such scenarios might be beneficial, that said, very theoretical.