... down to 6...the 'interpolation' reference i believe only kicks in in this specific mode.
Well, it's quite complicated stuff... Indeed, its not correct to say it's 6mp, because it's 2x6mp...
You should take apart an X10 raw file with DCRAW: you will find two frames, 6mp each. those are combined into one file, but the process to obtain a 4000x3000px file is indeed interpolation. In fact, you should imagine that sensor (or look at a good scheme) as two sensors on the same wafer, where the photosites of each one are alternate.
The NATIVE count of pixels is 6mp, but for two distinct frames that get combined into one image.
In case of 12mp use, the two frames are exposed equally, merged as median and upsampled at 4000x3000px. In case of 6mp, to obtain the DR expansion the frames are exposed differently (via an electronic shutter, yes), and merged in an HDR-style to get the best from each frame.
But you will read better explanations in the brother forum fujixspot, where there is a guy that explains things much better than I do.
Indeed, a 12mp image obtained by stacking/merging and upsampling two 6mp images is better than a simple enlargement, but not up to the detail of a native 4000x3000px sensor. Which is why the X20 has an edge in detail.
But yes again, you are right: it's inaccurate to reference the X10 as a 6mp camera, I simplified things too much. This doesn't change my conclusions, the X10 and X20 are different cameras, and the X20 is not simply better, 'cause there are things the X10 does better, in particular in the DR field; but only if making full use of the EXR capabilities, which happens at 6mp (two differently exposed frames stacked). DR expansion at 12mp is the same shittish method of increased iso/underexposed image we have in the X20.