It's not that simple. It's not like a) Some galaxies fit predictions, b) Some do not, and we add unknown dark matter.

It's that when they do not fit predictions, there is a single parameter (the mass of dark matter) that does explain its behavior, whose model is that of a weakly interacting gas halo. So dark matter tries to fix things in a very precise way, it's not an arbitrary curve fitting exercise.

Now whether there is really weakly interacting matter out there or some other explanation is valid is still an open question afaik; but the evidence is quite strong so far.