What I would suggest is ... “if it is XML, then treat it stem-to-stern as XML.”   Parse it using a tool like XML::Twig, and use XPath expressions to (effortlessly...) locate all of the <INFO> tags.   Remove the nodes, then transform back into text for printing.

Although this might sound like “extra work,” IMHO it really isn’t, because it thoroughly solves the problem, both in the short-run and in the future.   And it does so by pushing the hard work onto the backs of CPAN modules.