OK. I think there might be two things at work here.(posted 8089 days ago)
You say cutting the red channel reduces the dynamic range of your scanner? In that case, I suspect that Epson are guilty of using the infrared output of the scanner's fluorescent tube to gently 'fog' the CCD sensor, and so reduce the effective Dmax of the film. This is a cheap trick to make the dynamic range of the scanner seem greater, and I hope I'm wrong about it.
However, if that is the case; the infrared absorbtion of the film base, or residual AH dye, might be responsible for the mottling that you can see, but I must admit, it just looks like grain to me.
Do you know what the Dmax of your negs is? They shouldn't really be much more than 2.4D, and that density should be easily scannable without resorting to any IR trickiness.The second thing that might be responsible is just plain underexposure. TMZ 3200 isn't really 3200 ISO you know, it's more like 1200 ISO. Exposing it at 3200 will just put most of the image on the toe of the film's curve, and this'll naturally give some strange looking scans. Remember, a scanner is linear, and doesn't have the corresponding 'shoulder' of photographic paper to compensate for the toe response of the film. You should really be applying a curve to scans as a matter of course, to get the same result as a photographic print.
I might well be wrong on both counts above, and the problem could be due to something else entirely.
Anyway. Best of luck sorting out the problem Ryuji.