Because the people make the platform, and not the functions, and for lots of people you need a lower entry barrier, and the entry barrier for both of those is a good bit higher than fluxer.
Don't get me wrong, if matrix was a bit more convenient (easier to understand and to use like you would discord, and less bugs of which there are still a wide range of), I'd 100% advocate for it. But I can only tell my friends to use something if it's convenient enough that they will genuinely avoid a degraded experience.
Now this is a question: how far can you get with xmpp? Could you build an interface on top of it to look exactly like discord with all of it's functions? Or does something like that already exist?
My first instinct with these older protocols is that there's no way they could support 10 people in a voice call with concurrent camera streams and 3 screen captures. I'm genuinely curious how far xmpp goes.