You do not have to use transcription training, but you do have to edit the transcription within Scribe. To be clearer, once you select the audio file to transcribe, you need to "clean up" the transcription within Scribe. I've seen it improve tremendously over my time using it, so that now, lots of it is transcribing accurately, and when it isn't just right, the correct transcription is one of the choices in the Recognition section (to the right of the transcription).
So, at the risk of stating the obvious, if you copy and paste the rtf into another application and then make corrections, Scribe is not learning from that (don't be offended -- this actually didn't occur to me immediately in the haze of getting used to it initially).
Also, after I have done the final cleaning up of a transcription, I always have scribe do the vocabulary training of the file. (Note that it will not open .docx files, so save as .doc (funny since it requires the very latest version of Mac OS, but . . . )) I think this has helped improve accuracy too, since it is picking up jargon and acronyms correctly more often.
Choose Region
Australia, New Zealand, Asia, Pacific
Canada
United Kingdom
United States, Rest of the World