What other variables are there that actually matter and you'd expect to be different between the games Harper played and the ones he didn't? I can think of really only one - lefty starter. And that's not difficult to control for if one wants.
And as to the difference, well - .558 is a 90-win pace. .580 is 94. That's a difference that matters, and in the exact opposite direction you'd expect if you assume Harper is really, really good. Indeed, if you assume Harper is a 4-ish WAR player, it's literally the exact opposite in size.
I don't like going all stat nerd on people except when they make assertions about statistics that just don't make any sense. I'll stop now. To be clear, I don't really buy into that stat for some of the reasons we've both mentioned. I just find it a bit interesting.
I should have paid more attention during those grad level stats classes. 90 wins pace seem good to me.