... Just where does one draw the line between visible detail in both highlights and shadows and an acceptable and meaningful amount of detail in both highlights and shadows?
I've also noticed when DR is described in terms of decibels on spec sheets, it seems to be a lot higher than the generally accepted F stop range (6dB being equivalent to one stop - so I believe).
Yes, 6bD = 1 f-stop (because they measure in audio-inspired power levels instead of using voltages, which are proportional to luminance!). The overstatement you have noticed with those spec sheet measurements of "linear dynamic range" is largely because they measure the whole range from maximum white down to noise level, so at the bottom of that range, signal level equals noise level, and that can probably only be printed as either pure black or "mud".
So you get into the subjective question of how high the S/N ratio must be in a shadow region for it to be used to show details, or at least texture. Ray, didn't you find somewhere Kodak guidelines on imag quality in relation to S/N ratios, like 10:1 for acceptable, 40:1 for excellent? If so, please let us know where, I would love to read more!
Trusting those numbers for a minute suggests very roughly that about three to five f-stops at the bottom of the linear dynamic range are useless for the goal of recording shadows that can be brought up in level on a print (e.g. by "dodging") to show something interesting without just looking muddy.
Other glitches can come up too that all work to reduce effective dynamic range a bit: those spec sheet values are based on noise at an unilluminated pixel in a short exposure, and so ignore dark current and shot noise. Thus they are not the whole story for long exposures, and underestimate the noise level at pixels that receive a fair amount of light. Then there is the complicated debate about what sorts of noise can or cannot be corrected for in software ("fixed pattern noise" can maybe subtracted off, etc.)
