However, there is another side of the issue: utilizing some of those 2048 levels means at the same time increasing the number of levels at the low end.
What worth is an additional "low noise" stop, which contains only eight levels?
Using only the first 2048 levels means, that the seventh stop will consists of only FOUR levels; going to the right increases this to eight levels. So, what about the 70 levels?
There is an easy solution to solving the problem of excessive levels in the highlights and the paucity of tones in the shadows and that is to change the encoding of the data. That the brightest f/stop contains 2048 of the 4096 levels of a 12 bit capture is not due to the nature of light or peculiarities of digital capture, but to the use of linear integer encoding, which works well for current cameras, but would not be satisfactory when we get true HDR output from our cameras.
For example, in discussing the scRBG standard, [a href=\"http://www.anyhere.com/gward/hdrenc/hdr_encodings.html]Greg Ward[/url] has noted, "a linear ramp is employed to simplify graphics hardware and image-processing operations. However, a linear encoding spends most of its precision at the high end, where the eye can detect little difference in adjacent code values. Meanwhile, the low end is impoverished in such a way that the effective dynamic range of this format is only about 3.5 orders of magnitude, not really adequate from human perception standpoint...". You seem to regard plethora of highlight tones to be an advantage, but the experts seem to think otherwise.
By going to floating point encoding (there are special formats than represent the required precision with relatively few bits), the encoding precision is nearly constant over the entire range. Log encoding would make the precision totally uniform, but then you would no longer have raw data. These considerations are discussed by Ward and I would encourage thoughtful readers to look at his paper.
Since noise, not posterization, is most limiting in the shadows, no major manufacturer seems to have done this for still cameras. Another solution to improve shadow tones would be to add a few bits to the ADC, but thus far the results have not been impressive, again for the reasons cited above.