It's not that it's changing the lens characteristics, it's changing the way the photo is presented.
Extreme case: take a photo of a wall, process it in DPP from RAW twice.
First photo, set contrast to +5. Second photo, set contrast to -5. They are going to have vastly different vignetting *numbers*, because the numbers are taken from looking at the *photo*, and the photos look different even when taken with the same lens.
Now, i'm not saying that any review sites do that on purpose. But say one takes in raw and dpp processes at 0 contrast, and the other one just uses the camera jpg. The camera could be set to +1 contrast by default maybe? or even if the camera says '0 contrast' in its Picture Style menu, that could be more or less than what '0 contrast' in dpp gives you? Or even some thoughtless reviewer pressed for time accidentally leaves the 'vignetting correction' on by accident in-camera?
I don't know, it's all possible. In short, as long as you trust a single reviewer to keep doing the same thing over and over with the same equipment, then their results are comparable to each other. (Even if you trust a single reviewer, sometimes things out of their control mess up. Bryan at TDP noticed a bug in DPP a while back, where he'd set noise correction off for his processing, but in certain cases it turned itself back on and invalidated a lot of his results. At least he noticed and has since gone back and re-tested all the affected lenses, I hate to think how long that took him)
For further reading, here
is a well written article by Roger about comparing results between reviewers and sites. He's mainly talking about resolution numbers, but just susbtitute the word 'vignetting' and you'll still have the same idea.
Or read any review at photozone.de (at least in a Nikon review), you'll get the following warning:
We're performing our vignetting analysis based on (uncorrected) JPEGs straight from the camera. The JPG engine of the Nikon D3x features a rather flat gradation curve, thus has a moderate contrast characteristic, resulting in comparatively low vignetting figures - the corresponding Canon figures are roughly 40% higher due to the more aggressive default contrast setting.
Were any of those Sigma reviews taken on Nikons? Or even if they were all canons, do the 5D2, 5D3, 1Ds3, 1DX all behave exactly the same? (I think you get the point by now)