I try to avoid mixing Leiden system and lowercase when possible, but you probably don't have that choice here. I've just pulled a few books off my shelves that quote a lot of epigraphic or papyrological material in lowercase, and there appear to be three solutions. The first is to let them crash (although the book that did this also failed to align them with each other in spite of this, so I can't recommend it as a model). The second is to nudge them sideways - I saw this done with a gamma (Porson), but I didn't see an example of a dotted zeta or xi, where it would be more difficult, in that one. The third is to have all the dots much lower than the usual dotbelow is (as in Semitic or Indic transliteration) - I found this in a book published by the American Society of Papyrologists, who probably have to deal with this problem more than most. I didn't find an example of dropping them on the characters with descenders, but I don't have many papyrological books.

The biggest problems will be zeta, xi, iota subscripts, and maybe chi. My first inclination would be to go with very low dots, but that's probably not the easiest solution unless you're using a Greek face that has a some extra space under the descenders, and enough leading that it'll be clear which line they belong to.

Here is dotbelowcomb with a few greek characters in SBL Greek (a grab from FontForge).

Maybe John will like to comment.

Michel

I can't recall off hand whether I made the dot height contextual or not.

The connecting overline character used in Greek abbreviations is contextual in SBL Greek, so that if there is even one tall letter in a word the line is raised over all the letters. Obviously something similar could be done with the dot below, but I can't remember whether I implemented this in the current build of the font, or was waiting for feedback.

[Sorry, I'm very tired now, so can't bring myself to go and check. Feel free to download the font and test for yourself. Feedback much appreciated, here or in the SBL font user forum.]

Michael, What you show is basically what I've done. I felt uncomfortable dropping the dot lower on the gamma than, say, eta or mu, but it seems that's OK. It is not a convention I'm very familiar with.

Charles

I didn't find an example of dropping them on the characters with descenders,

Here is an example with descenders (coptic)

The example above is relevant provided the underdot you are dealing with is the one used to denote unsure reading. It is a mark that has nothing to do with the language and I then see no reason for it to look attached to the letter, like a diacritic. Here is a relevant comment from http://www.roman-emperors.org/debbiean.htm showing that the question is a bit tricky:

Another problem surfaced in a recent issue: How to handle marking the underdot for unsure reading, a common sign in epigraphy and papyrology. [...] Originally we had used the Unicode "combining underdot" for the unsure reading. However, the Unicode Technical Committee members have pointed out that it would be better to handle it differently, since the underdot has a specific phonetic meaning in many languages (e.g., a vocalic r in Sanskrit), but here indicates something different.

Keeping that mark low enough might resolve the ambiguity (it could even be found under an underdot!)

Michel

Keeping that mark low enough might resolve the ambiguity

But, depending on leading, it could be then confused with a "dot above"! ;-)