Argument from the extreme is a rhetorical fallacy. End of dicussion.
No model that is mistaken anywhere (extremes, normal or anywhere in between) is accurate.
These are the two main ideas on the topic, really. By my point of view, the following is the background for such statements. Correct me if I'm mistaken anywhere, as it is very much likely due to me not knowing very much about you and possibly to misinterpretations.
The first statement bases on the principle that reality in itself is too complex to be thoroughly described by models of any kind and shape. If it was possible to describe it in such a way, bringing a counter argument, be it extreme or common, would prove a logical fallacy in the model, thus invalidating it. By assuming that reality cannot be completely described, you are accepting to work with what is most suitable in the case, which is of course the majority of cases, excluding exceptions. This is acceptable, as you're going to miss a few cases with whatever model you are to use, so citing these cases won't prove any point.
The second statement, on the other hand, comes from the opposite principle: that is, reality can be described by models, which must be researched as thoroughly and tested as extensively as possible. Whether a final and conclusive model can be actually reached or not is another subject and doesn't matter here: both a Popperian search of impossible truth or a positivistic definitive approach have the search for a model in common. By accepting this search, you are accepting extreme cases as points to be subject to the model you're proposing, and therefore as possible arguments in the discussion.
