Yes. When the timestamps wrap it's fundamentally ambiguous, but this will be correct unless the timestamps are very far apart (and the failure mode is more benign: a really long time difference being considered shorter is better than all time differences being considered zero after the timestamp wraps).