Problem with sd1 Merrill and 10-20 4-5.6

Hi all, I just received a 10-20 4-5.6 and noticed an anomaly with the autofocus.

Normally I use only the central focus point. Well, with this lens set at short focal lengths, the autofocus works, but the accuracy is quite low, let's say +/- 2 mm on the rotating lens scale, in the sense that if I repeat the focusing it changes randomly with that error. Anyway it follows the distance, so that if I focus on a subject at 20 cm, it indeed focuses at about that distance but every time it changes randomly by about +/- 2 mm on the rotating lens scale. It beeps normally and it is normally fast.

To me it seems that the SD1 focusing sensor is not accurate enough, but this is a sensation. It is almost impossible to see the best focus by looking into the viewfinder with a 10 mm f 4 lens...

Due to the DOF, the inaccuracy is hardly observable in the center of the pictures, but at the perifery it makes a difference (experimentally, any idea why? Maybe perifery suffers of more aberrations, and adding defocusing we go over some perceptual threshold...).

At long focal lengths, say 20mm, the anomaly is quite reduced, but still there.

I tried different focusing points, but no effect.

My SD1 works perfectly with the 17-50.

The lens is now used only MF for distant subjects, the distance scale is spot on, and considering how these SWA are used, doing MF and DOF calculation seems the best approach.

I also noticed that the geometric distorsion is exactly zero at the focal length of about 11mm. Quite interesting.

I kindly ask to those of you having an SD1 and this 10-20 if they observe the same anomaly.

Thank you