Just a theory: say, if the core surface is not perfect, it acts like a lot of little 'needles' that apply a very localized pressure against copper. That means, both surfaces get more or less 'into' each other, to a point which amounts to the sum of irregularities heights divided by more than 2 (ie surface irregularities of HS and those of CPU core).
If as i suspect the CPU core material is much harder than copper (it's more like glass), and since most people (including myself) wiggle the HS/waterblock just a bit *after* mounting to squeeze more TIM out, i think then that the CPU core makes its way *into* the copper base... (for a short distance indeed)
And yes a little polishing makes that imprint disappear.
(edit) ah Bill our posts crossed each other
