Its better to say its model dependant.. String theory phenomonologists have to take a stringy inspired version of reality, make various choices, tune a bunch of parameters and throw in some extra string inspired structures (for instane Dbranes) and then they can indeed output various particle masses like neutrinos.
The problem is the choices are sort of inherently put in by hand to fit things, at a higher level. Which is fine, the standard model is no better (except that we had experiment to put things in by hand for us), but the problem is
1) Lots of presumably different ways you can do this
2) The phenomology is complicated, no one has ever gotten the vanilla standard model exactly, with everything just right, and no additional structure (which again is probably a good thing, b/c we hope to see more structure eventually). The problem is often this additional 'put in by hand' structure might lead to problems down the line, or it has various nagging finetuning problems, or it can violate astrophysics constraints, or it has hard to believe exotics, too many fcncs, etc etc. Its a monumentally difficult problem to solve, and no real first principle way on how to choose the best angle of attack. This is why some people believe breakthroughs have to first occur on the theory side, before progress in discarding models can realy take place.