On little indian architecture, the calculation of the correction field
was wrong in the PDelay_Resp_FollowUp message. This was introducing a
big random offset (few us) of the PPS.