As of the most recent (that we can tell) generation of entry-metadata, the entries for tagged terms contain text that is outside of the tagged term in xml: e.g., https://github.com/cu-mkp/m-k-manuscript-data/blob/4625a7ad236f37bca47f4bdc27fcd536e89bcbb0/ms-xml/tl/tlp003rpreTEI.xml#L19-L20
--> returns the phrases (see cell Y12 of csv) 1. subtly ground vermilion with and 2. <m><pa>walnut</pa> oil</m>, and if you add in a little (even though it catches nested tags, it then similarly adds the untagged text after the closed material tag)