When you have put a lot of ideas together to make an elaborate theory, you want to make sure, when explaining what it fits, that those things it fits are not just the things that gave you the idea for the theory; but that the finished theory makes something else come out right, in addition.

Well, at least, we want to make sure that the finished theory makes a testable prediction beyond the things that gave you the idea for the theory. (That property sort of corresponds to 'being theoretically progressive' in Lakatos' discussion of problemshifts, as discussed in chapters 3 and 5 of my book draft. If the prediction turns out to be correct, i.e., if the finished theory indeed makes something else come out right, in addition, that would sort of corresponds to 'being empirically progressive'.) If not, the 'finished theory' is just a description of the things that gave you the idea for the theory. This is the point that seems to be understood rather poorly in generative grammar. But come to think of it, that is understandable, given what seems to me to be the general failure of the field to recognize the significance of establishing repeatable phenomena and proceeding on the basis of them.