Cable skew is only part of the overall skew budget. Maybe it's easy to
make small, compared to the other contributors.
You say the spec is "significantly" smaller than the bit window. How
much? How much smaller than the signal risetime? (Sorry, pardon my
ignorance on this spec.)
Do you suppose they had higher bit rates in mind when they spec'd the
cable? It's good if your cable plant can survive a transition to the
next speed generation.
"But what harm would 90 degrees cause?" 3dB attenuation, for a
sinusoid. I don't know what the effect would be on ISI.
Is this the only pair in the cable? A skew causes more common-mode
noise, which might couple well between pairs.
If the wires of the pair were tightly coupled to one another, then the
characteristic impedance would vary significantly depending on the
balance of the signal (phase shift and amplitude match between true and
complement).
Regards,
Andy Ingraham