First off, the study period was 2002 - 2008. What constituted a TAA aircraft in 2002? A 2" black and white moving map? I also remember that the DA-40 was randomly excluded from the study because it was "too safe" and made the TAA numbers look too good. If this forum existed when the AI came out, the same discussion would be taking place with everyone saying the turn coordinator and whisky compass are a way better solution.
While I wouldn't be surprised if TAA is no safer than steam, when there is enough data to make an accurate conclusion, I'll be very surprised if synthetic vision does not have a meaningful positive impact on accidents caused by botched instrument approaches, CFT and maybe even loss of control.
As always, if you need an autopilot to fly safely you are not safe to fly.