To provide some practical reasoning behind this: ChromeVox and software like it, naturally, runs in the browser. A blind person is unlikely to have a visual impairment which means they can't use a browser but can use every other part of the system without speech feedback. Therefore, system-wide screen readers see much higher use - who wants to keep switching between screen readers all the time?

On the plus side, though, keyboard predictability breeds accessibility, so any testing in this area is often better than none.