Are manufacturers' Image Stabilisation ratings for zoom lenses deceptive?

I've been considering getting an L-series zoom lens, and the two that I am looking at are the Canon 24-105mm f/4L IS, and the 24-70mm f/2.8L.

Aside from the extra reach of the 24-105, the main difference is obviously the f/2.8 aperture versus the active Image Stabilisation.

Canon state that the IS on this lens "permits the use of shutter speeds up to three stops slower than normal". It's the "up to" in this statement that got me thinking about how image stabilisation is rated.

There is a general rule of thumb that your shots should be good (with a steady hand) at shutter speeds of up to 1/focal length, so clearly the focal length is a factor of image stability. Now, I can't imagine that Canon would publish anything but the best numbers for their spec sheets, so this leads me to assume that the "three stops" of IS they quote are at a focal length of 24mm. So my next question is, what does this mean at the 105mm end? I've done a few calculations, and I invite you to check my workings.

3 stops is a linear factor of 8 (2^3). I think this equates to a linear IS factor of 1.83 @105mm (8*24/105). Or, 0.87 stops. Less than 1 stop!

Using the same calculation but for a focal length of 70mm, the IS of the 24-105mm comes out at 1.46 stops.

Do I have my calculations correct?

PS: I know that the increased aperture of the f/2.8 will give greater background blur, better capture of moving subjects, blah, blah... so please don't feel that you have to lecture me on that. All I'm asking about is IS.

Answer

Reverting to some empirical data, the 70-200 f/2.8L II lens has a supposed 4 stops of IS at all focal length. DPreview tested it at 70mm and at 200mm and revealed it had just under 4 stops and 70mm and over 4 stops at 200mm!

http://www.dpreview.com/lensreviews/canon_70-200_2p8_is_usm_ii_c16/page5.asp

From the review:

We're used to seeing Canon's latest IS systems do well in our tests, and the 70-200mm F2.8 II doesn't buck that trend. Under these controlled conditions, the IS system is delivering something pretty close to the claimed 4 stops of stabilization (for example, at 70mm we get similar results at 1/5 sec with IS on as at 1/80 sec with it turned off), which is about as good as things get at the time of writing.

Unfortunately they haven't tested the 24-104 IS yet. But both their review and my experience with the 100 f/2.8L macro indicates the actual performance is pretty much in line with the official specs (again, four stops). So I'd be inclined to believe them and expect about three stops throughout the range on the 24-105.