i think people were overestimating the impact of integrated lle to users..
1) hle (which in pj64 has always been a mixture of hle and lle anyway.. we used to call it mle for medium level emulation

) is already so good, who's going to want the estimated athlon 6000+ it would need, for no improvement in most situations?
2) there are significant problems integrating low level graphics processing with hardware rendering. (and using software rendering to do all the texture filtering etc.? maybe double the requirements again to Athlon 12000+ or something). until someone solves the conversion problems, it can't look very good. may not be possible - because of course hle gives the author much more control to make the output in the form he needs. once you do something in lle, you lose control, and you can't have it both ways.
and other reasons.. i don't think lle is the holy grail people seem to think, for n64 anyway. this is also why it was added almost as an afterthought (hey look, we can do lle) and was never a main concern. if it had been, people would have been releasing this before 2006..
sorry if people were expecting lle=all compat problems solved and probably some day there will be an accurate all software emulator but it's not going to be something that's "better" for most people.
yes, this has happened mainly because donations motivated zilmar to do something fairly major in the rsp core, so thanks donators. i think he was initially just trying to fix some difficult games.