I'm aware of Arrow's Theorem, and that the only thing voting reform advocates can agree on is that plurality is the worst choice.
Most of the arguments come down to:
This weird thing can happen with the system I don't like.
This weird thing happened in these specific elections using this system.
Modeling a distribution of possible sets of voter preferences, my preferred system is less likely to do something weird.
I'd love to see analyses that measure what tends to happen across many actual elections -- in how many does each system reach the "wrong" result? (Of course, it's hard to gather data for elections that were run under plurality, because we don't know about second choices, approval thresholds, etc.)
All that said, I still think RCV is an improvement over plurality, and I'll support it for now, because:
It finally gives third parties a chance to find out if they're losing because (a) people are afraid to "waste" their vote or (b) their views are just not held by very many people,
Any crack in the nut of plurality is a win. Once we've tried one alternative, it will be easier to sell others. RCV is the thing with momentum now, but if STAR, approval, etc. take the lead later, I'm open to those as well.
On a related note, I strongly support abolishing the Electoral College.