Alastair M. Robinson wrote:
.ti3 file, profile and makefile here: http://www.blackfiveservices.co.uk/R300_Plain_20060122.zip
OK, taking the original sRGB IT8 chart as an example, the most saturated color in column 4 is J4, and this seems to be a target color of about Lab = 13, 32, -58.
Taking a look at the profile, the black (0,0,0) value seems to be Lab 39.388774 3.631250 -2.948365, so the black points are rather different.
Running xicclu on the profile:
xicclu -fif -ir -a R300_Plain_20060122.icm
and feeding in the target 13 32 -58 in in relative colorimetric we get:
13.000000 32.000000 -58.000000 [Lab] -> Lut -> 0.083257 0.000000 1.000000 [RGB] (clip) [Actual 47.500649 12.190514 -34.425555, deltaE 46.243540]
which is certainly a very large delta E, but notice that it's almost all blue, nothing else (8% Red).
Checking the B2A table:
xicclu -pl -fb -ir R300_Plain_20060122.icm
13.000000 32.000000 -58.000000 [Lab] -> Lut -> 0.090672 0.000000 1.000000 [RGB] so the same sort of value.
I'm not sure how any profile will be able to get a more saturated blue out of it than that really. Even 0 0 1 gives:
46.642085 8.602886 -35.679842 [Lab] which is a delta E of 46.6, which is greater.
Now your softproof indicates J4 at about 44 11 -22, which is a little different to 47.500649 12.190514 -34.425555, but it is in the ballpark, and it's a different exercise to figure out where these inaccuracies come from. (If I use your IT8 jpeg and do a relative colorimetric sRGB->R300_Plain_20060122->sRGB using cctiff, then J4 winds up with Lab 47 13 -34 which is a bit closer to what I'd expect.)
[Well, it's got rather late, so that's enough for one day!]
cheers,
Graeme Gill.