I hope all will forgive the long post, trying to both verify my understanding and perhaps help others by providing more detail.
1. Shot the CC with a D300 and D200 outside with cloud cover, exposures within seconds of each other, different lenses but both Nikkor.
2. Brought both nefs into ACR, white balanced (approx. 5500K) and equalized exposure using second neutral patch, applied new Adobe standard beta profiles to each. Very good match, first time I've ever seen truly equal images from the two cameras thru ACR.
3. Created DNG's of each cc image nef with DNG Converter.
4. Used DNG Profile Editor on the D300 image (white balanced), set base profile to Adobe standard beta, selected only the Red, Green, and Blue patches on second row. No changes to three colors, created recipe with "Edit both color tables smimultaneously" checked.
5. Opened recipe file with Wordpad, very interesting to examine, easy to see what's going on since in XML format. Six color adjustment sections, three for 2850 and three for 6500. Also the entire base profile in binary form is encapsulated in the XML file.
6. Opened D200 DNG file in PE, white balanced, used old 4.4 profile as base (deliberately trying to create different colors than D300). Loaded the D300 based recipe, exported new D200 profile.
7. Now back to ACR, loaded D300 cc image with Adobe standard beta profile selected and white balanced, and D200 cc image with old 4.4 profile selected and white balanced and exposure adjusted to match D300 image. Opened both in Photoshop CS3 in ProPhoto space. As expected, there was considerable difference in the color patches, both visually and by the numbers.
8. Back again to ACR, loaded D200 cc image with recipe modified profile from step 6 and white balanced and exposure adjusted to match D300. Brought into PS and compared to D300 image. Now the Red,Green,Blue patches matched well with the D300 image, only slight differences. Actually just selecting those three patches brought all 18 cc patches into a fairly close match.
In conclusion it appears to me that a recipe has enough information, with the color table entries and the embedded base profile, to "move" the same colors in another profile to achieve a good match. And it's apparently moving "both ends" (2850K and 6500K separate sections) of the profile appropriately, although I'm less sure of this. Quite an interesting and useful development by Adobe.