The loss of tables in PDF files was brought to your attention when you changed to the current (now old) method, yet nothing was done about it. The single column method at least gets us tables back in PDF's, but would it not have been easier to revert to the original (2013 or thereabouts) method?
Topic on Talk:Reading/Web/PDF Functionality/Flow
Table support got investigated (see e.g. https://phabricator.wikimedia.org/T135643 or https://phabricator.wikimedia.org/T73808 ). Some issues with the "original" method are listed in https://en.wikipedia.org/wiki/Wikipedia_talk:Offline_Content_Generator
True, this brings back the simple and working functionality there was some time ago. I do vividly recall, when I downloaded wiki docs as PDFs and I didn't have any missing fields or data! I don't know why they ever changed that service, at least now they realise that the current service is whack and needs replacing!!
If you want to "know why they ever changed that service", you could click the last link in my previous reply to read C. Scott Ananian's explanation why the "original" method had to get replaced.