This isn’t an answer to “what to do with how things are now”, but more of a comment, and also a question.
It wasn’t really the wordforms themselves that you wanted to delete. The list of wordforms shows what occurs in your texts, and those words are indeed still in your texts, so they should stay in the list. (It’s possible to delete wordforms with zero occurrences.) Instead, you just wanted to delete the analyses.
There is a command under Tools/Utilities to “delete all parser-approved analyses”. That would have gotten rid of anything suggested by the parser that you had not approved.
There is also a proposal (not implemented yet) to add a command for “delete all user-approved analyses”. That would have been the second part of what you needed. (Someone has written a FLExTools script to accomplish the same thing, but it hasn’t been published.)
I need to think harder about the state that you have gotten your project into—it surprises me that the Form column is not showing what is in the Baseline for your texts. Do you by any chance have more than one Vernacular Writing System? If so, it’s possible that column is set to “Best Vernacular” instead of to a specific Writing System, and that in one of your writing systems, the words are now blank, but in whatever Writing System is on your Baseline, the words are still there. And because the column is set to “Best …”, then you are getting a mixture of WSs showing in that column.
One way to test this would be to use Configure Columns to add another column for “Form”, and then set one Form column to one WS and the other Form column to the other WS, and see where you have blanks or not.
In order to figure out what to do for your current situation, we would need to understand what is actually currently happening. It’s a little puzzling to me.