Put in those terms, I can see your point, but because I prefer the varies ranking systems to FPtP I'm (perhaps unduly) attached to later no harm, because without a guarantee people giving ANY vote to a less preferred candidate won't cause your most preferred one to lose it'll be hard to sell people on it and we'll be stuck with FPtP forever. If I could pick any one of the systems I've seen proposed I'd probably go with Ranked Pairs, but it seems too complex to get popular approval (a smart chimp could follow IR and a lot of people STILL have trouble with it. ) Perhaps worse, it also doesn't guarantee you giving a lesser vote to someone you like less won't cost your favorite candidate the election. I can handle weird results as long as there's SOME logic to them, and if I'm following the graphs it seems like even in the most bizarre looking IR simulations IR a logic can be found.

For example, in the last (and weirdest looking) case there's a yellow triangle pretty much dead center of the red victory region. Since red has a greater y value and x value, if public opinion shifts from right of the triangle to the left it moves toward both while maintaining reds greater vertical distance; past a point yellows vertical advantage outweighs reds horizontal one and makes it the winner until public opinion is nearly even with red horizontally and red is again the winner. Effects of a vertical opinion shift are more pronounced because it increases the relative size of reds greater vertical distance, so less horizontal shift is generally necessary to make yellow the winner. The exception is that if the vertical position was always close to both red and yellow a horizontal shift left is more significant than an upward one and a win for yellow becomes a win for red. In all cases once public opinion gets VERY close to both vertical positions OR reds horizontal position red wins. The effect is a region of voters horizontally closer to red but occasionaly more concerned about reds greater vertical distance from yellow (despite being fairly distant from both themselves).

It's significant, I think, that in that simulation yellow has the LOWEST y value of the four AND is the closest candidate to red over all. It's a case where a single issue with strong public opinion can yield a surprising outcome, because if voters generally approve reds platform but are very extreme on that one issue red may not be close enough to their views, and they'll choose the most extreme candidate who's otherwise most like red. Red can only win if public opinion is VERY close to him on all other issues. In effect, it's a case where being moderate on EVERYTHING gets red beaten by a yellow candidate who's moderate on everything except the one issue on which the public is also extreme.

Maybe I'm fitting the data to my curve here; I can't be entirely sure, but that's my take: Just because the graph looks funny doesn't mean it's not representative, because electorates often look funny, too. Regardless, it looks like my choice ultimate choice is between systems that can penalize a candidate for shifting his position toward the majority or those that can penalize a voter for ranking candidates. Neither of those is ideal, obviously, but the former is less objectionable to me than the latter, particularly since it will take major public approval to implement either, and the general public will always be more concerned about hurting themselves than hurting their candidates.

After all that, I'm still very attached to mutual majority guarantees as well, because I still believe America's a center left country (at least in global terms) that's nonetheless had mostly center right and right Presidents since the war. To give you an idea of how strongly I feel about mutual majority, in 2000 I would've been happy with either Bradley or Gore (the Dem primary candidates), Nader (the Green nominee) or McCain (the GOP primary loser), but after six years of George Bush as my Governor the prospect of his presidency horrified me, yet the reality exceeded my worst nightmares. In my book, the Condorcet loser won, when a mutual majority system would've guaranteed me Nader, Gore or Bradely, all of whom would have eminently satisfied me. Given a choice between Condorcet completion or a mutual majority and later no harm guarantee, I'll miss monotonicity, but not as much as I'd miss the other two.