Heh, nice that they have plots of everything -- ESD waveforms, attenuation, eye diagrams!
Does show that it's not all that great, though. Peaks of 100-200V, and that will have a low Thevenin source resistance -- still dangerous to the transceiver. What ESD level that counts as (it may be closer to a CDM level condition!), and whether the PHY is rated to handle it, who knows.
If there are series resistors between CMESD and PHY, that'll help a lot; I'd be comfortable with it. With no resistors, I'd like something better.
Note that the CM filtering is very meager indeed, offering significant attenuation only from fractional GHz up. I never found much point in such things. But, maybe it's a good idea if you have a lot of radio modules nearby (cell, BT, WiFi, etc.)?
And going in the other direction, note that larger CMCs are straight out -- USB itself depends upon unbalanced symbols (J and K states) to frame packets. These get corrupted by CM filtering with lower cutoff frequencies. Anyway, there's very little impedance at the PHY for a CMC to work against (just the ~10pF pin capacitance), so the filtering obtained even from very large value CMCs is mediocre (even if it didn't screw up the framing).
In short, this is why USB is designed for, and must always be carried by, shielded media. The shield needs to be solid, contiguous between transceivers, at the very least well RF-grounded (i.e., bypassed with multiple capacitors, if not galvanically connected). (Galvanic grounding doesn't matter, as far as I've seen: the GND wire is tied to the shield on every port I've seen.)
Tim