That's what I said. But the calculation of a protein forming from the space of all proteins assumes they do randomly come together and join mechanically, so the calculation is the wrong one.
That does not appear to be the calculation that Axe was doing. What he did* was take a specific protein (not the naturally occuring sequence, but one intentionally selected to be sensitive to mutation) and then subjected it to random mutation to see how many possible changes led to a functional protein (his conclusion - few). He then extrapolated from this that the probability of any protein performing a
specific function was as low as 1 in 10^77.
Note that the probability of finding a protein which performs a specific function is of course a very different calculation than the probability of finding a protein with a function.
*with the caveat that this is what I understand he did - I am a little out of my depth.