I really like the delivery of more granular content objects for search results in this fashion - for too long we have got too hooked up on the concept of "search": the information retrieval cycle is one of moving from a requirement, finding (or "discovering" as Donna Maurer would put it) information fitting that requirement, and then - critically - making use of that information.
I'm sure I've not even got close to understanding what the CIS can deliver, but I like what I've seen so far...It fits nicely into the equation that I came across in Oracle Magazine (Zen and the Art of Information, George Demarest, p79-80 March/April 2005):
"Information=quality (data + metadata)"
I quite like this neat equation which expresses the idea that information is data with applied quality and context. Structured access to document sections certainly hits the context side of this equation on the head...