Similar to the comparison that you did of UNR vs Ubuntu MID, these results aren't very useful.
The Moblin folks have already pointed out, as you acknowledge toward the end of the article, that a comparative test between Moblin and Ubuntu would be unfair, given the debug settings used for the current Moblin.
Where the comparison of UNR and Ubuntu MID is concerned, Ubuntu MID is not a distribution that anyone would actually use on a netbook in its default configuration, which suggests that this comparison isn't all that useful either.
What I would be most interested in seeing is a comparison of performance (both CPU and battery life) between comparable i386 and lpia installations. With Ubuntu, to test a reasonable lpia install, you would have to start with either the MID release or the lpia alternate install, and then install the necessary packages to end up with an lpia config that is the equivalent of the UNR config.
Because they aren't only server-side. Firefox, Liferea are heavy users of sqlite.
Yes, that's true. The problem with the tests, is that they were all CPU/IO tests. There wasn't a test that indicates how well the graphics system was optimized. If all I were going to do was remote into a MID, those tests would be fine, but that's not exactly the point of a MID. There isn't anything that mentions battery life either.
The article doesn't really answer the question, "Which distro is best optimized for a MID?" There is a mismatch between the test-suite and the niche that MIDs fill.