Earlier this year, I ran a usability test of a standards-compliant E-commerce site with users of screen readers.
Results: Of two people surveyed, one could do nearly all tasks and the other could do none of them. What was the difference? Version of Jaws: Jaws 4.51 couldn’t do anything.
Feel free to refer to and cite this survey as a source. And if you want to know the backstory behind it, why, you could just ask.