@Robert: It could be the following bug I found while trying to work with "optical trains"!? Seems like the calculations of the focal length are done the wrong way around: