It's older than Descartes. It originates from the latin phrase exceptio probat regulam in casibus non exceptis. It has many ways it can be used, but I'm using it here to mean that the fact that there are only a very small number of deviations from the rule demonstrates that there is, in fact, a rule. If there were more deviations, we'd have to consider whether the rule is correct, but, because the number of deviations is small, that proves there is a rule.
Do you understand now?
What you are describing is rigid adherence to an insufficient model. The simple fact is, you are insisting on applying a binary model to a multi-state system. Just because the probability distributions of the multinomial favor two of the many possible states does not mean that the other probabilities are 0 or even negligible.