Upvoted, but I suspect this is a response to misunderstood or incorrect criticism, rather than a good meta-generalization.
All models are wrong, some models are useful. Outliers and exceptions are a reminder that the model is wrong—it’s not truth, it’s not reality, and it is not the ONLY thing available to make decisions on. AND an exception is not a knockout blow that invalidates all uses of the model. There may be MANY uses for that model, as long as it’s known not to be complete.