The way I read this test, it shows that "sometimes" Google gets original attribution right and other times it gets it wrong. I also see a lot of odd mixed results for very long tail searches - the kind you might do if you already know what content is available, but not the kind you would do if you are REALLY searching for something.
Google has already stated that they do not think always returning the original source is the best thing to do - that sometimes people want a more recognizable source even if it's only quoting the original. I'm not sure how I feel about that because I really can see both sides. Getting the message out there (the content itself) could be paramount in some situations, but certainly not in others. In others, getting the website the traffic is the key thing. It's a very complex issue, as I see it.