Transcription is omitting sections

I am finding lately that the transcription is cutting out certain sections of what I had dictated. It feels far less reliable than it used to. I think this behaviour has started to occur with the introduction of the faster model, but it is also happening even when I use the legacy models.