//urn:CustomerAccount/urn1:accountdescription = 'yoda'
Is really not the same as this query:
//*:CustomerAccount/*:accountdescription = 'yoda'
In Celtix ESB land, this last query (which is a valid XPath query) throws an error. It turns out the XML parser they are using does not support wildcards like this. Instead it will match any prefix as long as the prefix is presence. In other words, urn1:CustomerAccount and urn:CustomerAccount will both match the query. That's cool once you know it, but it is annoying when you don't. Dom4j is the culprit. No relation to Celtix but used in Mule which is used in the Celtix ESB.
You just have to do a little family tree tracing when using open-source solutions.