They may be reception/transmission times or PCR times. But there is a recommendation that the PCR times be used, as these are piece-wise linear and the stts table then compacts sensibly. The big question is, do we have a packet structure that allows the documentation of a reception/transmission offset from the PCR time?