I have added two more sets of words: the set of all single-word labels, and the same set but excluding zodiac labels. The two resulted to be quite similar, so I will only discuss "labels" in general. For these sets, I have considered the whole manuscript (while other sets only include pages with at least an image-break).
Like line-breaks, label statistics turn out to be somehow comparable with image-breaks.
For suffixes, line-breaks were a good match, with the exception of -am that is frequent at line breaks but rare at image-breaks. In labels, -am is much less frequent than at line-breaks; all the other values are reasonably close to those for image-breaks (though those for line-breaks are even better).
For prefixes, labels feature high frequencies of o- in all oGallows- forms. But, in the comparison with image-breaks, the values for o- are compensated by that for qo-. Image-breaks and labels are the only sets in which qo- is below 5%.
It should be noted that, even if both line-breaks and labels compare well with image-break values, line-breaks and labels behave quite differently: their similarity and difference with image-breaks are complementary.