If they were using any decent text to speech package that used the generic text to speech logic as long as no specific override existed for a particular word or phrase, issues like this would vanish quickly.
Having worked with IVR (Integrated Voice Response) systems in the past, this type of correction was commonly used when the generic text to speech came up with an atrociously mangled pronunciation of the desired reading. Mind you this type of solution was working back in the dark ages around the Y2K debacle when I contracted for the USPS (through NeXT / Apple of all people), and it was mostly corrections to city or street names, which is exactly the issue with the Model S navigation system.
Geez, I'm feeling like applying an old anti-Apple saying to Tesla: Bringing you yesterday's technology, today, at tomorrow's pricing! :redface: