This is the only reason to study STV, to look at the interesting patterns. Likewise, I'm fascinated by a knowledge-graph pattern (I'm not sure if it has a name) where we use something like k-means clustering to discover less popular candidates who belong to the same cluster of voters who voted for the top candidate. I think the results would deliver good government, in that it would surface the maximum number of non-extremist ideas. But I think the math is too complicated for most people to follow, and so I think it would lack legitimacy. So I don't bother recommending it, except as a thought experiment regarding alternate systems.