This isn't actually a question but a hint for others that might share the same problem: Parsing 1 GB+ large XML files and still be able to use DOM and XPath techniques. Source code (MIT-License) and description is available in my blog:
http://softwareengineeringcorner.blogspot.com/2012/01/conveniently-processing-large-xml-files.html
Hope this is useful to anyone!