If your problem boils down to this: " it still didn’t recognize the correct answers. Instead, every answer that it recorded was marked wrong", then that sounds like the same problem I ran into when rewriting an experiment in the new 2021 version: Correct response variable not recognised for new experiment with new version: 'key_resp.corr' always FALSE - #15 by TParsons.
The problem seems to have been corrected-just undocumented as a correction–in the latest version (2021.1.2). It took me some wrangling to get this new 2021 version to recognise the experiment I wrote in the first 2021 version (i.e., to update the “last run” python file), but it you play around with the version settings, you can eventually get it working. How, I can’t say, but if you mix it up and experiment enough, it’ll work.