The only thing we know for certain is that, if the Inmarsat data is correct, the aircraft is somewhere in about a million square kilometre area in the South Indian Ocean. Everything else has been an attempt to reduce that through clever maths to an area that can be searched for a price people are willing to pay. And the highest probability area was then searched. As I understand it, the area predicted by the drift models wasn't far from it.
Where I do somewhat disagree with the ATSB is that there were a few likely candidates that weren't made high priority search areas. I remember early on that someone pointed out you could hit all the arcs at a constant speed if you assumed the autopilot had been set to 180 degrees after the final turn. That would have given a relatively small area to search just to see if it was there.
I understand their reasoning for not doing that, and wanting to take a systematic approach, but I would have looked there if I was in charge.