The problem is that it depends on how you assess it. There are two, both perfectly valid, ways to look at this.
The way you're looking at it is you see it as a state-only contest. B got the fewest votes, B's votes go to C, C wins the state. The end. From an administrative level this is the simplest approach. I don't feel any need to expand on this assessment as you're in favor of it and seem to grok the principles behind it.
The other, equally correct, way to look at it is to assess this as a national contest. In that case, C is the one that actually gets the fewest votes because they have 0 electoral votes in any other state. C is incapable of winning, so C would be eliminated in the first round of the state level contest. After all, one of the points of RCV is to eliminate the impact of spoiler candidates that cannot win. With that in mind, it'd be dumb to design an RCV system that increases the impact of a spoiler candidate that cannot win.
The issue with the first interpretation is the risk of magnifying the impact of a spoiler candidate who cannot win. The issue with the second interpretation is the sheer administrative difficulty (if C were competitive in multiple states then each state needs to take into account how other states are doing their RCV process, etc.). Both flaws would be unlikely to matter >99% of the time, but that one time the flaw would matter could lead to a constitutional crisis or less dangerously result in fundamental dislike of RCV systems. That one time might also become more likely if voters feel more comfortable voting for third parties due to this system.
The problem here is that both systems are fair, logical, and valid; they also each present major issues in edge scenarios. That's why it's important to go for a popular vote first. That way the election is one election, which RCV is explicitly designed for. The current two layer 51-elections that lead to another election that we have for the US presidency is basically a nightmare scenario for an effective RCV system.