yea, the transcription of the spelling out of the name is great, works as intended. for instance if I say, "My name is Gregg spelt g r e g g" it'll 100% of the time transcribe as "My name is Greg spelt g r e g g" but it'll have my name in the structured data be: Greg not Gregg as the spelling would show it to be. but from what i'm understanding, since the transcription is more or less correct (at least with spelling out my name) but since the data is wrong it means that I should focus on the prompt BUT if it was the transcript that was incorrect, I should upgrade the model? is that right?