My naive understanding is that there are so many tunable parameters in string theories that you can use them to model our universe but not to make predictions outside what our existing models already predict. So you can falsify a given set of parameters, but if you find a set of parameters that fits our universe there's no way to know if those are the right parameters or why these parameters here and not those parameters there that also fit.
At any rate, a "theory" that can't make predictions is no theory at all, in my book. And a theory that encompasses quantum gravity should make predictions that are more precise than any previously made.
I'd expect such predictions to be untestable in practice, using contemporary methods - the theory as expounded should encompass any test that can currently be performed, without the need for predictions. Otherwise the theory is falsified at birth.