Yes, I agree that Einstein's special relativity does almost seem to be a deduction from observation, for precisely the reasons you give. Maybe I can shelter my earlier remark under its "rarely if ever" qualifier. This might constitute the "rarely". Einstein did have to embrace a hugely counter-intuitive assumption though, even if it was kind of hanging right there in front of everyone in experimental results like Michelson and Morley's. Nobody else seemed willing to grasp it.
I'm inclined to lean towards methodological and ontological pluralism at the moment.
I like to think that scientists have a whole toolkit of assorted methods that they dip into as needs dictate. There isn't any established algorithm that tells them which methodological procedure or tool to employ when. Part of scientific creativity is figuring out new and original ways of deploying existing methods and maybe even inventing a new one now and then.
And I'm inclined to think that there are different kinds of being. Physical objects exist in one way. The laws of physics exist in another way. Fictional characters like Sherlock Holmes exist in yet another way. We need to make sense of ideas and word meanings. Numbers and the abstract structures of mathematics certainly seem to have some kind of objective existence. We probably need to have some account of unrealized possibilities which might arguably even be observed having physical effects in quantum mechanics. The past, present and future seem to have different kinds of reality. If it is to have any claim to completeness, metaphysics will have to provide an account of all of that. (I expect that some of them might be reducible to others.)
Maybe my pervasive pluralism rubs off from living in the vicinity of Stanford. (For those who don't recognize the reference, I'm referring to the so-called 'Stanford School of the philosophy of science'.)
http://philosophy.ucsc.edu/news-events/colloquia-conferences/stanfordschoolphilosophyofscience.html