Hi Gary - you've discovered the "edge of the XML universe". This is not to do with ROC or NK, instead its that the world of "XML processing" has an explicit assumption that the processing context is the entire document.
There's also an assumption (I guess from the SGML days before XML) that the document schema is unique and often is actually tied directly to the document (eg DTD, or XSD declaration - make my teeth itch).
In the compositional relative ROC universe this can be very frustrating.
Due to the global document context assumption, you won't find an off the shelf XML validation technology that can cope with validating fragments of a document embedded inside a compositional structure. This has never been part of the design scope of standard tools (because of the explicit assumption they all adopt).
However all is not lost. Schematron is the most flexible extrinsic validation technology - and its use of XPath with assertion is a very nice approach. It also allows the path to be easily extended to apply the same sub-path to validate a compositional fragment. Have you tried this? It also has the added advantage of being much easier for developers to write and maintain.
PS Thanks for alerting me to the URL exception in Jing - this is another bugbear. The XML (and RDF) toolset only know the basic set of URL schemes and (since they're validators) get nasty when its something local. They don't provide easy extension either - so we'll probably have to go in and hand override the class that thats doing this to get it to play nicely. Can you please send me the full stack trace so we can find which class is getting upset.