I was recently thinking about this very topic and I agree with your sentiments, although, I'm not sure if your graph or any other visual representation or verbal description can truly do the idea justice. The sample size of any graph isn't really important for this idea. What is important is that +EV graphs are very incomplete pieces of information and there are a lot of ways to "run good" or "run bad". Those +EV graphs are just the easiest stat within arm's reach.

How you run when the money gets all-in IS obviously important, but so are a lot of things. How often are you getting good hands like AA and KK? How often are your good hands running against other good, 2nd best hands like when you have AA and they have KK? Or, you flop the nut flush and they flop the 2nd nuts? Or you get top set and they get 2nd set? How often are you hitting the flop or your draws? We've all gone through periods of getting AK and getting nothing time and time again and having to fold. Or your suited connectors neither suit nor connect, but sometimes you run good and they do. What about table selection? Sometimes you get a drooler just giving his money away and you run bad and can't capitalize, but sometimes you run good and do. Sometimes you don't get it all-in before the river, but plan it street by street so you WILL be all-in by the river and villain sucks out on the river to beat you, but you have too much invested to fold and their range is wide enough to commit with worse, so it's correct to finish your plan but you still lose and it doesn't show up in your EV graph that you lost or you won if you got lucky.

I recently had a 3 way pot all-in against a short-stack for 30bb and against a full stack for 100bb. I had KK versus the ss's AA but the full stack had KQs. I statistically would have netted 40bb after losing to the ss but winning against the full stack had the %'s held up, but mister KQs luckboxed a flush to beat us both. In HEM, I was a 9% favorite, but realistically I was a 80% favorite or so to net money.

All-in EV is not a very useful stat EXCEPT if you have a large enough sample size and could prove that you are getting your money in good more often than bad, that's usually an indicator of a good player.