My overall sense is that the underlying reason for this is that supporting math in typesetting is an extremely difficult problem, much more so than people realize. Most typefaces don't come anywhere near being adequate for math, and those that are functionally ok are often really not very good aesthetically speaking. The equation typesetting can also be really difficult, in terms of "under the hood" implementation issues as well as the specification of standards. etc. etc. etc.

As a result, it's not so much LaTeX is good, as much as it is that it's been around for so long, and the domain is so much of a slog, that it's hard to develop something better.

I think a lot of people have a limited vision of what things could be like, and tend to base impressions on limited experiences with alternatives. The proprietary nature of Word does create problems in sharing written documents sometimes; the "black box" nature of WYSIWYG programs in general can be infuriating when you need to be careful with typesetting equations. On the other hand, constantly having every typesetting detail in your face with most LaTeX implementations can be intrusive and leads to "forest for tree" problems sometimes.

There are lots of beautiful texts typeset in LaTeX. But there's also a lot of LaTeX documents out there containing correctly typeset equations that are nonetheless really unpleasant to read or look at, for visual design reasons.