My understanding is that the FADEC system should have prevented an overtorque
A few reasons, that I would suggest. One is that the FADEC can only react to reasonably collective rates, and a very large and abrupt pull with always cause a transient overtorque until the FADEC 'gets its act together.' Another reason might be that the torque limiting is 'soft' compared to other parameter limiting, so that in attempting to restrict to 120% one FADEC allowed its engine to rise to 121.5%. I've been told from the horse's mouth that the S92 Tq limiting is not particularly 'hard,' and we can see that in practice during training. A further explanation might be that the FADEC was limiting N1 rather than Tq, and therefore as the Nr decayed (if indeed it did) the torque rose.