The approach described doesn't use a StaX parser, and probably isn't going to be as robust to failure or as extensible as using a StaX parser. If you look at the code, my patch allows you to specify the XML element name, "namespace_prefix", and namespace_URI when identifying the correct tag. My patch also makes it easier to massage the XML too when reading in data.
Initially when I tried to create a XML parser, I tried to hack something up like the previous approach described. But after trying to parse real-world data (e.g. a dump of wikipedia), I threw up my arms and decided to use a proper pull-parser.