I am working on an LTR transliteration of an RTL project. For some reason, the directionality of punctuation around numerals in a PDF export is sometimes reversed.
Everything looks fine in Paratext, e.g.:
But in the PDF export it looks like this:
Sometimes it is okay, but it is often very messy.
I thought there might be some invisible direction markers from the RTL source text still hanging around, but I have searched for U+200E and U+200F; they are not there. The numerals and punctuation seem to be standard encodings.
I do not observe this behavior in “plain” LTR projects, just transliterated ones. Is there some setting that might be left over from the source project that I need to change?