A model which is not validated by experiments is nothing but a pretty picture, maybe useful to adorn a PhD Thesis or to impress an ignorant audience.

However, the mental models we produce, whether they are structural models, models of biochemical processes or any kind of scientific hypotheses can strongly influence on the way we think about a problem

Therefore, despite the limited precision of protein structure prediction, a good structural model of the molecule can be an extremely useful tool for the experimental biochemist as it helps to formulate questions and to interprete experimental results.