If I have a document in "ProfileA", which is completely within ProPhoto RGB, and I convert this document using absolute colorimetric to ProPhoto RGB, and I convert this using absolute colorimetric back to "ProfileA", shouldn't the LAB values remain the same throughout the same process? (Except for possible rounding issues, which I'm willing to consider negligible, especially in 16bit color mode.) Isn't the whole point of absolute colorimetric that in gamut values don't change? Everything starts out in gamut in "ProfileA", and since ProPhoto RGB fully contains "ProfileA", then everything should be in gamut there, and likewise back again to "ProfileA"?
I have a color in "ProfileA" of RGB 35/43/35 LAB 4/-3/3. Converting that to ProPhoto RGB, it becomes RGB 24/25/24 LAB 13/-2/2. (RGB of course changes - it's the LAB change that confuses me.) Converting it back to "ProfileA", it becomes RGB 53/57/55 LAB 7/-4/3. I'm sort of willing to overlook a change of 1, although in 16bit color, I wouldn't expect it often. But where did L* changing from 4 to 13 to 7 come into play?
Hard to say, but do both profiles use the same illuminant (D50)? Any Chromatic adaptation?
Which color engine is used for the conversion?
I saw in another thread that you used Adobe ACE, have you tried another (e.g. Microsoft or Argyll)?