It isn't too clear from the description, but isn't non-saturation of the inertial measurement responsible, rather than saturation? I'd expect that some computat

It isn't too clear from the description, but isn't non-saturation of the inertial measurement responsible, rather than saturation? I'd expect that some computation did not saturate its output, and the result overflowed (and hence became negative), which in turn was fed into altitude measurement.