If you only think of the election in the terms that RCV brings forward, then by definition all RCV elections find the correct winner. The Burlington RCV election essentially disagrees with two other ways of determining the winner of an election, and likely it would have disagreed with two other methods. If you look at this website, which compares voting methods using the same election, you'll find that RCV (listed as IRV) is usually in the dissenting opinion as to who should be the winner. If you play around with this spacial simulator you'll find that, not only can you generate nonsensical graphs with RCV (showing win scenarios had just plain shouldn't happen) but they take longer to calculate, too.