I have created a bug report with Adobe (FRMAKER-9478 ) for this bug. I have copied the problem description from there for reference.
When F_ApiGetTextForRange and F_ApiGetText return text containing mdash and ndash characters, with the flags FTI_String | FTI_TextObjId and encoding set to FrameRoman, they are returned as \x14 and \x13, instead of \xd1 and \xd0. If the flag is FTI_String, they are returned correctly as \xd1 and \xd0.
The bug occurs with both 2019 and 2020, but not in 2017. The bug report lists workarounds, the use of UTF-8 encoding is the one I have implemented. Please note that other special characters may be affected and/or other FDK calls. I have verified the two mentioned above fail.
I would appreciate it if you would vote for this rather bizarre bug to be fixed by Adobe.