Color Management - Am I Doing Something Wrong

Question

I am encountering a problem that I see a great deal of discussion about but to which I am not finding a solution (if there is one).  After converting a jpeg to sRGB, I am seeing oversaturation, particularly in reds, in some applications and on a digital photo frame.
So far this has cropped up in two scanned negatives that I did not have tagged from the scanner, moved into Lightroom (as a 16 bit TIFF), moved into CS5, converted to ProPhoto RGB, and edited.  I then flattened, converted to sRGB (trying it 3 alternative ways, e.g from the LR export module, from PS edit>convert menu, & from the Save for Web and Devices Module).  On my color managed applications the colors look fine.  I first noticed this in the file thumbnails for the images and then when viewed on a digital frame.  Also in some applications (but not others) I see this oversaturation.  The only workaround I have found is to assign my monitor profile to the image.
I do not have a wide gamut monitor (HPLP2475W), although I think that would create the opposite problem, and I reprofiled my monitor with an Eye One after this problem first appeared.
After trying many different things, searching around on the web and rereading my color management stuff, I still haven't figured out where I have gone wrong.  Any ideas?  The only hint I have as to what may be wrong is that in soft proofing, my Monitor RGB seems to be much more saturated than the image as it appears outside of soft proofing.  As I indicated, I reprofiled my monitor but that phenomenon remains.

Answer 1

Based on what you have said above and after hitting the books, this may be where I have gone "wrong."  Let me know what you think.  I am scanning color negatives and as such cannot create a custom profile for the scanner output.  (At least that is the conventional wisdom from everthing I have read, mainly because of the number of variables that could make a profile next to useless).  I brought the image into Lightroom with the full gamut of the scanner but without any tag or embedded profile.  I then exported them to CS5, but I probably should not have converted to ProPhoto RGB at that point because, as you note, I was "approving" its appearance at that point even though the image had no starting profile to begin with and except for a global Lightroom adjustment, I had done nothing to correct it.  Thus, I am assuming that what I did was to start down the color management path with an incorrect assumption about the accuracy of the color values in the image from the beginning.  And that would cause my output at the end after conversion to sRGB to look wrong on an unmanaged sRGB device.  Do I have that right?
Instead, I probably should have done any of the following:
(1)  Using Nikon Scan (also an option in SilverFast), let the scanner software profile an image in my chosen color space (as opposed to the full color gamut of the scanner) as best it can, and move forward from there.  Nikon Scan uses a one size fits all canned scanner profile and then (presumably) converts it to the color space of choice.  Silverfast must do essentially the same thing for color negatives but, as a starting point, gives you profiles for various kinds of negatives, with the caveat that it also cannot hit the accuracy mark dead on because of the variability of negatives.  Either of these options present the easiest, least challenging, but perhaps suboptimal, solution.
(2)  Bringing the untagged image into the CS5 Working Color Space from Lightroom and then previewing the image under different profiles until one is found that comes closet to matching the expected colors and gamma of the image, and assign that profile.  Then edit the image from there, converting as necessary.
(3)  Leaving the image unmanaged, but in the working color space handed off from Lightroom (ProPhoto RGB), manually color correcting without the aid of a starting profile, and only converting to a desired color space (e.g. sRGB) for a specific output device (e.g., a photo frame or the web).
(4)  Assigning the working space handed off from Lightroom (ProPhoto RGB), manually color correcting without the aid of a starting profile, and converting to other color spaces as necessary for a specific output.
I'm not sure there is a huge amount of difference between #3 and #4.
Do these sound like the correct options to you?
Thanks,
Brian