Your response, I think, well illustrates my point. If your output is abstracts, reviews, and papers then you are not documenting as I have narrowly defined it above - you are working on materials for your own circle.
I think I'm realizing that the problem here is one of responsibility. To use your terminology, you are claiming that it is the responsibility of the researcher to document everything he (or she, whatever) does for every "circle", and I'm claiming to the contrary.
To that end, let me describe the firehose that is today's science publication. It's been two weeks since I last read through my science RSS feeds, which contain selected publications from about ten journals. In that time, there are around 1000 entries. That's just a small selection of the field... there are many more journals I'm choosing to ignore simply so that my sips from the firehose aren't completely overwhelming.
I will suggest that having every researcher document every finding for every circle is a tremendous waste of time, as the vast majority of that "documentation" would never be read, as no one in the outer circles has interest in it, even if it was written in an accessible manner. In that mindset, I don't think it's the responsibility of the researcher to document his work for the public. Furthermore, given the huge amount of research put out each week (like my RSS feed above shows), just surveying all of it regularly to find what's interesting to the "outer circles" and then writing up in appropriate terminology is close to a full-time job.
I know that in neuroscience, the field I'm currently working, we have the Society for Neuroscience (http://www.sfn.org/) which puts in a lot of work to make the general findings accessible to the public. They do a yeoman's job, I think, and if you browse their site you'll find a lot of very recent science written for the "outer circles", as you call them. If you're interested in a field, find that field's SfN and check out what they wrote; don't look to the experts themselves for regular updates.