Thanks Ian; ok, I think I see what's going on. It turns out that we are both doing exactly the same thing (before you "correct"). My method finds the the lens extension (ie., how far the lens needs to be "racked out" from infinity focus) by trial and error. This insures that the distances are accurate.

I first thought perhaps you were adding on 2.6% to roughly correct for this. But it appears the "+1" in your formula actually does the correction. Both of our methods give the exact same result. So the only error should be due to the location of the lens nodal points (same as principal points here).

That said, I think that using the 2.6% factor is not the best way to correct for this. A single distance adjustment would be better. I once had some Rodenstock data sheets showing locations of the principal points, but not to be found. (The current website data also does not show principal points.) Just for fun, I thought I would check a Rodenstock lens - it's a 75mm Apo Rodagon D, for 1:1 reproduction.

It appears that my Rodagon D has overlapping principal points - that is, it should make images in LESS distance than the calcs indicate. Since you had found the opposite, I thought I'd test a bit further. I set up for 1:1 reproduction, which requires that the subject and object be 4 focal lengths apart, plus the correction for principal point offsets. Actual results were that my lens did it in about 3/4 inch LESS than 4 focal lengths, which is the same as the estimated overlap of principal points. So it is not a given that more distance is always required.

I know that few people care about this sort of thing, but just wanted to point out that the two calculation methods give the same result, and that the only error it might have is due to the (generally) unknown location of the lens' principle points.

ps: the linked diagram shows standard notation for the principal points, H and H_prime.