The 18 megapixels are indeed strange, because they match a 6,8µm pixel-pitch. But the new CCD-architecture by Kodak has a 6µm pixel-pitch which would result in 24MP. Always remember: we're talking about pixel-pitch, not pixel size! As I was told by Kodak, the actual sensitive area of the pixels has barely changed over the last architecture-generations (9µm, 6,8µm, 6µm), only the gaps between those areas became smaller!
The 6,8µm-architecture was introduced in 2004 and the new generation (introduced with the H3DII-50 and now S2 with microlenses) seems to be superior to the previous generation regarding noise and DR (just like DALSA).
There are lots of strange comparisons Digital vs. Film, most of them with horrible film-processing. Some claim a 35mm Velvia is similar to their 11MP 1ds, some compare it to their 6MP D100 and other claim 3MP as digital aquivalent...
Velvia has 80lp/mm at a realistic contrast of 1:1,6 which would result in about 24MP at 35mm. But film also has grain and contrast slowly decreases with higher frequencies and additionally has to be multiplied with the scanner-MTF. But from my own experience, clearly visible >10MP are not a serious problem with well-processed 35mm-scans (which would result in about 25MP in 645 and 45MP in 6x7).
I've tried it carefully with my M8, the files are clean but don't contain more information than a properly done, oversampled, denoised and sharpened professional scan. Occassionally I still like to put some Velvia/Ektar (Dynamic Range of up to 15 stops!) in my MP and pay for the scan, but mostly I stay with B&W (Imagelink HQ, only 25ASA but bloody sharp!)
Here's a properly scanned 6x7 slide (even when f16 might cause slight diffraction and Provia isn't as sharp as Velvia!):
http://www.rockgarden.net/download/60MP_from_6x7/Try to get that amount of detail with any 35mm-digital-solution...