My employer is moving from trans to MQ's VR platform - Bilboa

My employer use MedQuist’s DocQScribe software and is moving us all onto its VR.  So far, what we’re seeing from the VR is almost completely garbage, but MedQuist claims that our corrections will “teach” the VR to recognize our dictators’ speech patterns.  Of course, our wages are being renegotiated, as well.  We have about three dozen MTs and I think a few hundred dictators. 

So far, it takes me about 25% longer to correct this VR trash than it would take to delete-all and transcribe from scratch (but of course that’s not allowed).  Admittedly, though, I’m totally new to VR correction.

So, questions for everyone here with experience of both, especially those with experience on MedQuist’s DocQSpeech (I think that’s what they call it):

• What percentage increase in productivity would you expect (on average) once we’re all accustomed to editing VR?  (This is our HUGE question, because we’re trying to estimate how many of us will be laid off – not to mention what we should expect/bargain for [we’re union, at least] as a line rate.)
• How much time should we expect for our own “learning curves”?
• My own experience is that, focusing as intently as I possibly can, my accuracy on VR is still well below what I can do transcribing.  Is this normal?
• Advice? 

 Thanks!

-- Bilboa