Natural selection can most definetly lead to useless proteins.
during DNA replication, errors can occur. these lead to deviations from the original specimen. these deviations can be one of three types:
negative; the deviation harms the creatures chance at life, as it hinders its chances at getting food, escaping predators etc.
positive; the deviation gives the creature a better edge as far as survival
and last but not least, deviations that have no real bearing on the ability of the creature. its is neither useful nor hindering. however, as, over generations, these deviations 'stack up' on each other, it can lead to greater developments.
so evolutionary developments do not neccessarily have to pass through 'useful' stages
Right, but the point is that the development of a complex system which can only be developed through a long sequence of useless stages would be vanishingly improbable, so while possible, we would not expect to see such a history of development. If such a system existed, it would indeed be fair to call it "irreducibly complex", and it would indeed be good evidence against the theory of evolution. However, many of the systems which so called "ID scientists" call irreducible complexity are not, and the arguments they make that something is irreducibly complex are unscientific, and generally proceed by
argument from ignorance, which is not a logically valid argument. It may be that an "ID scientist" cannot imagine how a structure may have been formed in a series of small steps, each of which confer a survival benefit, but such a sequence did occur. To truly show irreducible complexity, a scientist would have to do better than that - they would have to give a logical argument that such a sequence of events could not have occurred.
So far, nobody has done this. Indeed, there is a great deal of evidence for evolution (although not all of the details have yet been explained), but no concrete evidence showing it did not occur. In fact, I would go so far as to say that the only reasons for disbelieving the theory are religious, rather than rational.