So far I have only tried Nikon D300 profiles, for example "Nikon D300 Camera Standard v4.dcp". And the command lines are of the form
dcptool profilename.dcp test.xml
dcptool -c test.xml newprofilename.dcp
One would expect profilename.dcp and newprofilename.dcp to be equivalent. And until now they have been, otherwise dcptool would be useless.
I guess there are in the new profiles some previously unused tags, which Lightroom and ACR understand but DNGPE and dcptool do not (yet). Hopefully those tags are documented in the DNG spec. DNG is an open standard, right?
More food for thought:
Besides Lightroom and some Nikon DSLRs, I also have PSE9, CS4+ACR 5.6 and a Nikon 1 V1, which was unsupported until LR 3.6 / ACR 6.6. Now, after the new camera profiles have been installed (they are common to all Adobe products), PSE9 and ACR5.6 both recognize Nikon 1. As they should. But camera specific profiles produce strange looking images. Same thing with v4 Nikon DSLR profiles. Adobe standard is fine.
My conclusion is that there is some secret sauce in those new profiles, recipe so far known only to new LR and ACR. Open standard?
And dcptool cannot fully decompile a profile, it only shows those parts it knows of. I think.
Just idle curiosity on my part, I like Adobe Standard. But it would be nice to know the facts.
Yes, well, there are indeed new tags. Notably the following, from exiftool:
Exif 0xc7a4 : 1
Exif 0xc7a5 : -0.5
Exif 0xc7a6 : 1
I'd guess these are offsets to the curves, or something similar, to get round posterization issues. That's consistent with what a V4 profile stripped of the new tags does, which is basically to overexpose.
However, the bad news is, until Adobe publish a new version of the SDK and documentation, dcpTool will remain broken for V4 profiles. I'll try to do an update of dcpTool once the new SDK is out. Well, soon after anyway - getting the SDK and the XMP toolkit to compile is always a nightmare.