This is just a wider reflection of a very common pattern--success is only vaguely correlated to quality. If anything, success is heavily dependent on qualities of the environment rather than the thing in question, so there is probably no way to just look at something in a vacuum and decide if it will succeed. Really, the metrics aren't at necessarily at fault, the "market" is--people choose an inferior product for whatever reason and then stick to it.

You can see a division like this in many other fields, particularly the arts and literature--most popular literature isn't "great" and most "great" literature isn't (as) popular. So the natural parallel is that PHP, Python...etc are like the thrillers at the top of the best-sellers list and Scheme, Haskell are like what you would read in an English class.

And really, this makes sense--whenever anybody talks about the quality of a programming language, they are talking about whether they would use it themselves rather than whether the public at large would use it. So it is completely reasonable to have a "lower quality" language be more popular than a "higher quality" one.

Coincidentally, while I talk about high quality and low quality, I do not mean to denigrate popular languages. After all, sometimes a thriller is all I want on a plane ride! And it could be a perfectly fine book indeed. But that does not mean it's a better book than Ulysses.