[Spliff]

I understand that SQLite's FTS probably switches to all-lowercase storage, and so, "text" export should be retrieved not from that FTS, but from ("stripped") rtf? (The other alternative would be to store THREE versions of the content in UR: the rtf, the FTS, AND the "text", with correct = original lowercase / uppercase: that would be probably "insane" though...)

I cannot say HOW / WHEN it's that way, but I'm absolutely positive, having observed that phenomenon many times: Some "text" export is all-lowercase (except perhaps then for some "proper names"!), whilst other (i.e. "plain") "text" export is correct, i.e. preserves the original cases for the whole text.

From my point of view, this is totally unpredictable - I admit I haven't spend additional hours into further analysis -, so for the time being, when I need "text" export", I use "rtf" export instead, which works faultlessly (!; I then use "¬" as the necessary "page" / item separator), then open in TextMaker (I have a paid license (as well as for MS Word but which is much too much of a "fuss"), but their FREE version is as good as the paid one for this task - I checked that!), save "as text" from there, and then run my AHK scripts onto that TextMaker "plain text" output: works perfectly, so anybody encountering that problem will get a free, and robust solution to it.

On the other hand, it's quite weird that UR's "text" output switches between plain text with correct case, and - unusable - almost-all-lowercase - perhaps depending on the "original origins" of those contents? (I had imported many UR contents from other sources / applications, and then, lots of my UR contents are "originally UR" - perhaps that element may explain it? I really don't know...)