Welcome to the Q&A Forum

This is a good read. http://www.seomoz.org/blog/duplicate-content-in-a-post-panda-world I think you should be careful with robot.txt because blocking access to the bot will not cause them to remove the content from their index. They will simply include a message saying not quite sure what's on this page.. I would use noindex to clear out the index first before attempting robot.txt exclusion.

"There is an XML sitemap submitted and GWMT shows a total number of indexed pages in the 800'000 region." Brilliant. That's the number I would trust. Incidentally, I see different numbers than what you see for all 4 site: queries you mentioned. Variances are pretty normal in my experience. I've never noticed it, I would be intrigued to hear if someone else has correlated such variances to a technical issue or penalty.