Bearfoil, mm43's backtrack calculations and plots are solid data as far as source data can be found. Indications are that it may be fairly accurate as a guess, since more a priori logical places for impact have been searched and found wanting.
Losing the VS at altitude can partially explain why the plane was not out where it was expected to be found. I'm not sure about that. But I can grant it. It does not seem to explain the apparent attitude and impact vector when the plane hits the water. And a complete break up at altitude does not seem to explain some of the damage observed such as that on the crews rest facility.
I'll grant it's a valid potential to look at. But an event tree from altitude to water that avoids "and magic happens" and too many "lots of choices here that make no serious difference would be helpful. The events on the way down need to be 'dense enough' to explain why 121.5 MHz remained silent, for example.
That is one of the great unexplained mysteries of the flight. Why, as soon as the change over to the African station could not take place, didn't the crew contact a nearby aircraft on 121.5 MHz to ask them for contact help? That should be SOP, I'd think. If events happened so densely that coping with them left no time for the radio it starts to make sense. But that would mean the problems began far earlier than any of us have suggested so far.
So the best tools at present are inventing hypothetical event trees from normal flight to impact with the ocean in roughly, within a few miles, of where mm43's backtracking put it or at least into an unsearched area. (At least the engines should have gone straight down unless they did not separate from the rest of the plane AND the fuel tank bladders did not lose integrity.)
I'd suspect that is the sort of investigation BEA is struggling through at this time. With a little thinking you can easily see how many branches each second of the way down might have. At least there are rough event "checkpoints" on the way down that might be helpful, the ACARS messages and the gap in the messages come into play as points that can rule out large groups of branches in the event trees. But the problem is that we don't even know within a gross guess how the plane broke up, into many large pieces with a lot of small pieces or all small pieces.
So give some slack, presume that BEA is doing this (at least for exercise), and then start doing it ourselves somehow. When we note a hypothesis starts to break down, "file a bug report" and let people try to solve the "bug". The sad thing is that while we are motivated here many of us have day jobs, too, I suppose. (I am playing hooky right now myself.)
{^_^}