Not defending Behe, but it sounds like "impossible to have evolved" is a decent definition of "irreducibly complex". So far, that's a legitimate topic of discussion. The question becomes, "is there such a thing as 'irreducible complexity' in biology?" It may well be possible to answer that question one way or the other--for example, to identify a genetic structure with the property that no mutation is viable, or some such thing.
The part where Behe begs the question is when he demonstrates that something is mighty complex, and concludes that it's irreducibly complex. He's equivocating on the definition of "irreducibly" and taking advantage of the fact that one possible meaning of the word is "super-duper".
He defines as "irreducibly complex" any biological system that requires substantially all of its parts to function. Take away any part any small subset of parts, and it ceases to function.
Behe makes the falsifiable claim that such a system cannot have evolved in a Darwinian fashion.
Unfortunately for Behe, his claim has been falsified on numerous occasions. That is, it has been demonstrated that certain irreducibly complex systems, according to his own definition, have evolved in a Darwinian fashion (eg. Mammalian blood clotting systems and the E. Coli lactose metabolic pathway).
There are several problems with Behe. The first is that it is impossible to define irreducible complexity in an objective way. I suppose an atomic particle is irreducible in the sense that anything simpler would not be functionally equivalent.
But complex systems are almost always comprised of parts that are functional in themselves. Sometimes the parts were originally used for some unrelated purpose.
Behe's examples have pretty much expired with time and further reserch. They are not irreducible. The direction of findings is not in his favor.