I've been having similar problems with a site of mine. Apparently, Rogerbot does not follow the same rules as Google's spiders and, as a result, Roger should be taken as a rough guide instead of a definitive answer.
I know, this isn't a particularly helpful answer. If anyone could chime in with the specifics of what Roger does and does not crawl and how it relates to google's spider crawls I'd appreciate it as much as you.