UPDATE Well, I managed to get 6500K white when I created a new document in Photoshop CS5. So, I think my profile is OK. But there are some software related "issues". My previous test image was a white rectangle which I created in MS Paint (Win7 x64 version) and saved in PNG format. As much as I know, MS paint is not an ICC aware software. It measured ~5900K in paint. When I opened this PNG image in Photoshop or IrfanView (CMS plugin ON), I measured ~5940K. So, I assumed that "something happened". When I used Photoshop to converted this image into the device profile and saved it as a PNG format, spotread measured ~5640K on the IrfanView (CMS plugin OFF) window. And it measured the same on the Photoshop windows, as PS used the device profile for rendering. (If I enabled the IrfanView CMS plugin on the converted PNG, I got a false result. A blue image. But this is OK, because this ismage was already converted and tagged with a simple sRGB profile, not the device profile...) When I opened the unconverted PNG (which I created in PS) in IrfanView (CMS ON), I measured ~5940K again. Firefox 4.0b6 hehaves like IrfanView. (May be PS tagged the file with a V4 profile which is not supported by these softwares. Or.. I don't know...) So, the profile works in Photoshop, when the image is correctly tagged. But this is not enough for me because I want proper colors in Firefox AND MPC-HC. MPC-HC shows the same behaviour as IrfanView and Firefox. It does "something" but not what I want. The image is processed because I can see a yellowish but relatively smooth gray gradient instead of the uncorrected bluish grays + yellowish 100% white. So, I think that the MPC-HC implementation is broken. It uses littleCMS 2 which theoretically supports V4 profiles as well. But there are no profiles here. It works from the device profile and user selectable presets (desired gamma and rendering intent ; and an input standard preset which can be auto,Rec709,Rec608, etc). But it has to be enough because the source standard is defined. It knows that Rec709 has D65 white and sRGB primaries. And the tonal response curve is an open question in the "community" but the picked up one of the possibilities (flat gamma 2.35, I think) but this won't affect the white point adaption (which is D65 for sure...). May be I should talk with the MPC team about this. But I am not absolutely sure about this because it shows the same behavior that "many" other softwares. I also played with the MS Photo Viewer -> same as IrfanView or Firefox. But it theoretically supports V4 profiles as well. I never measured 6500K here. The only situation where I achieved the "correct" result is the new document in Photoshop. (May be I forgot to check the PS saved image in MS Photo Viewer.) Any ideas? PS.: Does the -u switch in colorpro do anything with YXZ cLUT + matrix profiles? (And can it play any part here? Or any other switches...?) I am wondering what would happen if I get an ICC file for Rec709 and I create a Perceptual B2A table with colorprof. May be MPC-HC would use it when I set it to Perceptual intent mode. But it looks like it works without the tables. (May be it creates it's own tables from the colorimetric data.)