The PR version 23.4.0 now has Text-based editing. There are two types of transcription; source media (adapts to changes in the portion of a clip used in a sequence) and static (does not change).
The source media transcription does not allow choosing the audio track, and always works on Audio Track 1. See this post:
Also, it appears to me, without any Adobe corroboration, that the actual language pack files may be updated at any time. And there are multiple files for each language. But the accuracy generally should always be better.
Either of those issues might affect accuracy of transcription.