I don't know where you're getting this from. Dark matter models make lots of testable predictions, and they currently match the overall data set (which includes many different observations, not just the galaxy rotation curves that are most often mentioned) better than any other models.
> not to mention stuff like string-theory
Here I agree: string theory has not made a single testable prediction.
At this point I see string theory as more of a mathematical framework than physics
No, it didn't. Particular supersymmetric extensions to the Standard Model made testable predictions of what particles the LHC should detect at different energy levels, none of which have panned out.
String theorists like to take credit for those predictions (or at least they did until the predictions became notorious for their wrongness) because they claim that string theory is what justifies those particular supersymmetric extensions to the Standard Model. That doesn't mean we should let them get away with it.