This is a follow-up issue to
the xref repair algorithm simply searches for the nearest offset, which may fail if more than one xref table is present
Once we have a sample pdf which can't be parsed with the simple algorithm, we can open a new issue.
And here's one:
That file does have more than one xref table.