My training - start at the furthest point that an item could have reached and then work in.
If you are looking for something on the surface that is moving, that is a good strategy. You trap the item between its last known position and the furthest possible point and the search area should shrink over time as you clear sectors.
For an item on the bottom that is not moving, it may not be a good strategy.
The furthest possible position is a guess. The search area should not vary over time (except as new information comes to light) and the LKP is a good starting point (assuming it is reliably known).
Of course, the retrospective analysis approach can always find the flaw in the grand plan that should have been recognized at the beginning.
They did find AF447's resting place though and that fact should be respected.