As an intel guy I guess I'm of two minds on this issue and I've seen similar debates take place within the intelligence world. One of my jobs used to involve trying to get intelligence downgraded and/or released for the same reasons Cavguy lists for pushing info down to the unclass level - access. Information is of no value if the people who need it don't have access. In the intelligence world, a careful weighing takes place and information is often "sanitized" to remove source or whatever it is that keeps the info at the higher level. There are procedures and policies to accomplish this as well as personnel who are trained to do it.

One thing that concerns me with the unclassified caveats is that it's not clear (to me at least) who adjudicates whether something is sensitive or not. Without some standard and without knowledgeable people weighing the costs and benefits of placing information in the classified, unclass or FOUO categories, there is likely to be inconsistency in how the caveats are applied. Another problem, which will hopefully be corrected through standardization, is that too often a document is simply labeled FOUO without any idea of what parts are actually sensitive. Paragraph markings are a good thing. Finally, there doesn't seem to be an expiration for FOUO unlike actual classified information. At what point does a FOUO document become simply unclassified with no restriction? Who knows. That's a problem.