How could we approach the Pharma and Recipes?

[MarcoP]

Looking more carefully at the lexicon in BNF Lat.6741, it turns out that the order is not strictly alphabetical: entries are grouped by the first character, but the other characters appear not to be considered. Also, there are entries with the same initial word, where the second word is an adjective restricting the meaning. For instance these are consecutive entries in You are not allowed to view links. Register or Login to view.:
Caprifoliu(m)
Creta alba
Creta viridis
Cristicula
Cerulens
Carbo
This illustrates both the repetition of the first word and the fact that only the first character is used for sorting.

You are not allowed to view links. Register or Login to view. are about different plants ("herba morella", "herba sandix", "herba vactimum", "herba viola").

There are two different stones ("lapis") in 9v, two "niger" and two "nigri" in 10r, two "rubeas" in 12r, two "succus" in 13r, You are not allowed to view links. Register or Login to view., two "viridis" in 14v.

In the end, the rate of repeating paragraph-initial words is not dramatically far from that in Q20.

Also, as in the BNF ms , some of the repetitions in Q20 cluster in the same page. For instance, four polaiin paragraphs appear in You are not allowed to view links. Register or Login to view. (You are not allowed to view links. Register or Login to view.). There are a few other cases, though less extensive I think. Of course, this evidence is not enough to claim that Q20 is semi-alphabetically sorted, but the differences from the colour-lexicon seem to be fewer than I initially thought.

I must say I am happy with these findings: thanks again to Nick and Anton for sharing their ideas!