It seems that there might be quite a difference between how the 9000 and V (50) scan Kodachrome.
I read all sorts of arguments all over about V/5000 vs. 9000 for slides, some saying that the 9000 blurs things and is too soft while others insisting that the V/5000 with there direct, non-diffused, lighting emphasize grain/scratches/defects too much.
Anyway, I got to see a demonstration (using default settings and no great care taken on only a single slide, keep that in mind- with care and optimal settings or using alternate software supporting more control it might be closer) of V vs. 9000 for a Kodachrome slide.
It almost seemed like the V (and probably 5000 as well) maybe really does over-emphasize grain/scratches/odd defects. It's almost like the harsh direct light is glinting off everything at a 45 degree angle and making stuff show that maybe really shouldn't, at least not to that degree? I noticed that if you looked at the original slide from an angle right under a direct light you could see a lot more defects than if you looked straight through it head on.
It still may be that the 9000 merely blurs or misses things, but I'm leaning toward it being that it actually captures things in a more natural and true way. The V scan also showed a few weird signs of stair-stepping along dark skinny lines against brights that the 9000 scan was free of.
The V scan didn't seem to handle sharpening in photoshop very well without doing fancy NR first with grain, scratches, etc. quickly getting out of control. Once again do keep in mind it was not scanned in an optimal way though!
If you had large bright regions next to ultra deep shadows I could swear it almost seemed like a lens flare sort of contrast fade out swept over the dark areas with the V. I guess it was probably just due to lower contrast settings/poor default settings or something. Then again I did read some complaints about bright bleeding into black if they were right next to each other. However, I expect they were only talking about a pixel or two distance bleed over and not a wide flare effect. Anyway, whatever the reason, real and troublesome or nothing at all, I didn't see that on the 9000 scan.
The 9000 scan also showed a lot better shadow detail (if also some speckly noise not present with the V, an 8x multi-scan got rid of all the noise though and kept all the detail) and richer color with the default settings, however it also seemed to expose things a bit brighter so it might not have been a fair match. Plus software such as silverfast unlocks double exposure and multi-scanning options in the V which the nikonscan (which was used) locks off. I suspect that Silverfast double exposure might get it close to the 9000.