Thanks for the replies. I should have clarified how I scan in my original post. I am scanning with an Epson V700 flatbed, at 16-bit, 4800 dpi, RGB, with all corrections turned off, linear capture curve, and full contrast range.
Archiving
I archive the negative scan at full resolution RGB in jpg2000 format, between 80-300MB usually.
Working Images
For working purposes I convert to grayscale tiff (sGray), with several non-destructive adjustment layers for cleaning, correcting contrast, and adjusting shadow detail. These negatives are destined for Blu-ray, so I have chosen sGray as a close match to the gamma of the Blu-ray colour space. It's not exact, but close enough for my purposes.
The green channel is typically sharpest (I always check), so when I convert to grayscale I discard the red and blue channels. Usually I down sample to 2400dpi for the working images. Just before being sent to Premiere, where these images will become part of an audio-visual, I generate a third file, a 1080P jpg.
It's a long complicated process. Each image takes about an hour to scan, archive, edit, index, and create a version for Premiere. Then after all that, maybe another 30 minutes per image getting it into, and manipulating it inside Premiere.
It's the valley, not the humps
Another thing I should have made clear: it's the valley I'm interested in finding out about, not the two humps. Why are a large number of levels effectively missing from these negatives? What's the cause? I've rarely, if ever seen a valley in scanned slides.
This is not a question about how I fix the valley, or about scanning. I'd just like to know what causes that valley. Probably the only people who might be able to explain these valleys are those who have been extensively involved in B&W developing and printing. To me, it seems that a chemical explanation is required.