I read those articles and didn't see anything supporting this claim. I agree a 6850 will crossfire with a 6870, that's not what I'm questioning.
In "effect" in load terms yes but crossfire now, unlike in the past does not auto adjust clock speeds, if you take a 4870 and a 4850 as an example for how it used to be (if I remember rightly they were clock adjusted), both cards would run at the default 4850 (slower) speed, unless of course the user manually adjusted clocks so they were both the same (if that was possible).
Today, if you run a 5870 and a 5850, if they were reference the 5870 will run at 850, the 5850 will run at 725 (of course without manual adjustment), that, as you have pointed out may well lead to one GPU being more loaded than the other, simply because it is slower and therefore having to work harder and/or as you mentioned...... the SP difference? However the benefit now for enthusiasts is that they no longer feel that they are losing that overclock ability in their cards, they can run 2 different models at max overclocks (say 1050mhz for 5870 and 1000 for 5850) and see the performance benefits
I could not find this quote anywhere in those articles. It's actually a TPU forum post. What I can't find is a reputable source or even a tech site that backs up your claims.
Legit reviews did an article about 5870+5850. Their GPU-Z tab showed the cards at separate clocks (850 and 725), however in performance the 5870+5850 performed at or below the 5850CF (within 1-3 fps). The article also states:
One thing that shocked me was that the CrossFire gaming performance of a Radeon HD 5870 + 5850 was lower than that of a pair of Radeon HD 5850 video cards in CrossFire in all of our testing with the exception of one game. I would assume it is because the higher capacity card (the Radeon HD 5870) gets brought down to the lower one (the Raeon HD 5850) plus some overhead. This basically goes to show that while mixing and matching video cards to run CrossFire works for better performance it isn't the route to take for peak performance. If you want the most from a CrossFire configuration it is clear that you should pair two identical cards together.
I would like it to be true since there would likely be a performance boost if it worked. However, everything I've found says mixing cards in CF still results in forcing the clocks to match.