That's an example. One can argue about whether the "bad" candidate should have actually won in that scenario, but what's strange is when you have a situation where the "good" candidate would have won if only some group of voters hadn't put them first.

I think at the root of it is that removing the candidate with the least number of first-place votes is kind of an arbitrary way of eliminating candidates. Maybe that candidate should have been eliminated, but maybe they were an ideal compromise candidate, tolerated by most voters even if not their first choice.

Perhaps RCV might be significantly improved if one were to eliminate by borda count instead of least-first-place-votes, but that's not the system that's been advocated widely in recent years.