I appreciate your suggestion, especially since it fixes the problem ;-)

Yes, the $p->unbroken_text(1) solves the problem... and I made a note about moving the processing. The "skip" is fine because each input file follows a hardcoded structure.

As a matter of fact, after reading some snippets here, I re-wrote the code using HTML::TokeParser and it works perfectly and it's much simpler this way :-)