Hi, I developed a perl script along the lines, that I described earlier,
(and some shell script for wrapping it,)
and I created a CSV file from your PDF file.

Are you interested in that (CSV+perl+shell -- only a 1st shot) ?
Let me know!

I leave the heuristics to "human intervention".

What do I mean by that?
The scripts lists per page and per document all the possible "physical" column numbers,
that you might want to specify.
Some of them do not really makes sense from the human perspective.
You tell the script at the 2nd run, which ones you regard as meaningful,
and the script creates "logical" columns for you.

Yes, this needs a more detailed description, I agree.
And maybe one by a native English speaker.