Each of the models we have built so far is incapable of explaining the behavior of more than a small part of Nature. Most don't even do that very well—they are usually only "good enough" for the limited purpose to which they are put.
Nature has so many interactions at so many levels in so many dimensions that no conceivable model is capable of describing them all. The best scientists can do is simplify the problem to the point where the resulting model is only valid in a very limited domain.
Even if a one-model "theory of everything" could be constructed, it would be so involved that no person could make any use of it for purposes of prediction. Among other things, execution of such a model would require more resources than are available in the observable universe. It would be far faster simply to observe results in Nature.
_____
The really unfortunate part of this is that professional scientists tend to get emotionally attached to their models, ignoring or discounting any data that suggests the model might not describe all it is said to describe.
