I suspect we will mostly agree when we drill down on terminology, but to make a point:

"The whole point of real science is to find the model that best accounts experimental measurements and better predicts future ones."

I can create a model that predicts that every time a magician reaches into his hat, he pulls out a rabbit.

The model can be insanely accurate, matching past data and predicting future rabbit pulls, still just having a model like that is an intellectual dead end.

A model is nice and all, but what we really need is an explanation of how the hell the magician is doing it. We need an explanation for what is going on "in reality". 

Once we start conjecturing explanations for that, and critiquing them, that will lead us to more problems and questions, which will further expand our knowledge.