to Jim L about shooting in different spaces, the logic is in there you just have to sort it out.
When you create an image in a space [gamut] and define the parameters of that space, the damage or deletion is permanent: You cannot put back what was removed. There is an overlap of color range gamut to gamut, but the definition of volume of the profile is limited to each individual space. Now, there is an sRGB content within the Adobe RGB space so when you shift back to Adobe RGB from sRGB the volume is adjusted, spread and the respective contrast and personality of the Adobe RGB will be recognized, the color is not fully complete as it was in the original file. The larger gamut is a much more sensitive space so overall quality will prevail.
Using my analogy with the three vessels: If you take the cake pan [Adobe RGB] and pour it into the 16 oz glass [sRGB], 48oz of information will spill down the sides of the glass to never be seen again. Then take the same 16oz and pour into the quart jar [cmyk] and it is only half full, but Photoshop (or whatever application used) will try and adjust this content to fit the profile as best it can using the information supplied. Obviously, the more information provided, the selection process is going to be optimized and the profile will be more absolute. When you take the 16oz and poor back into the Adobe RGB cake pan, the information that is supplied will assimilate, providing accurate compensation for fulfilling the profiles definitions - but now you are short the original 48oz of pure data. However to your surprise, you will see a shift in contrast, saturation, luminosity, and brilliance, but it will never be as good as the original file.
Keep in mind though, to go from ICC profile to ICC profile in the cmyk gamut is minimal in shift by comparison to the drastic loss going from a non-defined RGB, or Adobe RGB, down to sRGB; this is a major data shift. If you go from RGB to cmyk and back to RGB one, two, or even three times, there is a minimal loss of information that a printing device can reveal as noticeable, but do these conversions too many times and color will begin to block.
In conclusion, we always save the original as a master file so if a new output is necessary it will come from a parent file, not a child. My workflow is something like this:
1. Original capture > output .psd file (parent file)
2. Color correct, clean, retouch, and finish file keeping all layers - Master file saved as .psd with either non [specific] assigned RGB or Adobe RGB (still parent)
3. Flatten file and assign profile for specific output needs - saved as Tiff usually (first offspring)
4. If jpeg or www file is needed, resample down flattened Tiff, save as jpeg (more than enough information going from the older sibling to the newborn)
If I am asked to reproduce a new file with a new ICC profile, Colormatch or generic RGB space, I do it from my Master file.
Hope that helps,