101 almost invariably indicates that the subject was speaking or some other noise was occurring at the start of the voice RT. DirectRT will not accept any voice RT under 100ms. Consequently, if there is sufficient sound at the start of an RT, DirectRT will accept it as a response at 101ms from stimulus onset.

Any response time greater than 101 indicates that there was not sufficient sound at the start of the trial to trigger an RT, indicating that the response was probably not a carry over from the previous trial. To obtain an RT of 300ms by way of a carry over response, the subject would have to be silent on the current trial until 300ms into it, and then resume their response to the previous trial. I suppose this could also happen if their speaking volume on the previous response dropped below the RT criterion until 300ms into the next trial. The only way I can imagine teasing that apart from a valid response to the current trial would be to manually check the sound recordings.