Lol. I had typed up a full response and then inadvertently closed the window o_O
What Pfeil said, BUT, turning the listening on and off may not be enough time (it was my first thought, too). My suggestion would be to put a one-second pause right before the start dictation action (which is what I'm doing here and it seems to solve it for my setup... you can try varying that to a half second to see if it works).
What is happening is that the speech engine is picking up one long stream of what it thinks is speech from your speakers. The dictation mode is turned on before there is sufficient time for the current, 'speech' to be ignored, so, it's getting picked up as dictation. You know that brief pause you have to do between commands? It's just like that.
Wondering if there is something I could do like what I had done with the, 'Reject Pending Speech' option but just when dictation is turned on... ?
Hope that helps & welcome back, Jerry!