Quite right. This second experiment presumably measures time-of-flight in the same way as the first one did - and I'm betting there's an incorrect assumption (or perhaps subtle equipment malfunction) in that measurement.
I think I recall reading that they didn't take the orbital motion of the GPS satellites into account, with respect to the difference in inertial reference frame of the GPS clock and the surface clock. They apparently calculated as if the GPS clocks were on the ground, not in orbit.
Ah yes, here we are:
Faster-Than-Light Neutrinos Result of GPS Movement
The motion, once factored in, came to almost exactly the time difference they measured in the neutrino arrival.