Like others said, you do get better temps with direct die, but only 3-4C better at 1.3v, and I agree with Apariton, not worth the hassle for 3-4C.
As for surface area question....If you have a soldered IHS with 87 w/mk intel indium solder and mx2 (5 w/mk) paste between ihs to waterblock, and you delid the soldered IHS, and then use only mx2 paste between die and waterblock, then surface area plays a role, and you get worse temps. Because the heat spread from small hot spots on die to large IHS size (spreading hot spots out as well) at 87 w/mk speed, before coming into roadblock at 5 w/mk prior to delidding, and after delidding the situation is worse by replacing tim1 with paste 5 w/mk when surface area to transfer heat was so small. So when comparing different thermal interface materials for tim1, you have to account for surface area.
However if you have liquid metal between die and ihs. And then you remove the ihs and use the same liquid metal between die and waterblock, clearly removing the ihs will give better temps, because the same thermal interface material was used, so surface area is effectively cancelled from the equation.