DISCLAIMER: I have no experience with XML::XPath, the following comment is based purely on my reading of the XML::XPath docs.

I think maybe that statement in the docs is a little over zealous. I believe the suggest is being made with the intent being that you shouldn't construct a new XML::XPath object on the same file/data over and over each time you want to "find" something new -- that instead you should make one XPath object for each XML datasource, and use it over and over.

If you have two seperate xml files, then it seems like you should definitely have two seperate XML::XPath objects.