lets think critically here.
what is the use of TIM?
1, to transfer heat through surfaces.
2, to fill gaps between to surface.
as such, you want the material with the highest heat transferance, covering as much of the surface area as possible. however since even the highest heat transferral is still less than metal itself, ideally we want to have 'just enough' TIM without having any reducing returns.
on the CPU heatsink, heat is localized in the center yes, but due to thermal conductivity, the whole surface will be transferring heat. As such you don't want to limit yourself of results by forgetting to fill the gaps on the edge, which btw, has the most gaps. (as most heatsinks are very slightly, convex.
you can also prevent 'over supply' of TIM by having a very tight mount, excess will be squeezed out. (of course, not telling you to silly butter the CPU.)
as such, the thin film method should scientifically give the best results, given your mount is tight enough. both blob and cross methods runs the risk of having heat sink surface not fully covered, (and you won't know, unless you unhinged it, which will introduce too many air bubbles and not recommended.) As such, the thin-film method also proves to be the savest bet, and I expect it to give the most consistence of performance on a tight mount.. and that was what I found to be true.
food for thought.