Standard practice is to look at a bunch of data that do not make sense, and to decide that it would be easier to come up with an abstract conceptual model, rather than actually focusing on the anomalies. Once the model is built, model data become the pseudo-explanandum, which generally prove the theory pretty well, since the data were generated by the theory. Then, any real data that happen to coincide with model data can be cited as further proof of the theory, and it all wraps up into a neat little package. But there isn't actually any value in that, because it didn't yield any information that wasn't already there. When such work gets accepted, and it becomes difficult to challenge, it's actually negative value, because it is precluding a more accurate description of the phenomenon. This has been going on for so long that now, there is an enormous amount of potential value awaiting us, if we are willing to see past the accepted constructs, and build more accurate theories that can actually predict future events, instead of just explaining away existing data.
Modern science has become very abstract.