The article looks like it could be interesting, but without the actual XML to play with, it really doesn't give me anything to test or try.  A picture of part of the text doesn't give us anywhere to go.  Also, you mention that your method might not scale well, but don't expand on it.  Given the people who already posted on this thread, they may be able to help refine the technique so it will scale well.

BTW, I am your target audience.  I'm very fluent in T-SQL, but not in parsing XML.  I'm also the type of person to tear something down to its wires to figure out how it works under the hood.  Your approach looks intriguing.  Could you post the full XML?