I have to question this test because I can draw the test pattern more accurately myself in a different app (Draw!). So something is obviously off.
I downloaded the app they use (DrawNoteK) and Draw!(Free). After a few quick tests, it was very apparent that there are significant differences between the two programs. Draw was noticeably better at drawing than DrawNoteK.
DrawNoteK's performance dropped off significantly as I drew faster. Draw represented curves more accurately, while DrawNoteK displayed curved lines as a series of sharp angles. DrawNoteK also had slightly more lag. Download both apps and try it yourself.
So I would like to see the test re-run using Draw! instead. I would also like to see the recording in real-time (not sped up). Results will probably still be inferior to an iPhone, but not as horrible as indicated by the test.
Also, the squiggly line down the left side of the screen almost looks as if the "finger" was straddling the edge of the sensor. So occasionally it read as off-screen, and occasionally read as on-screen. This would interpolate out to a squiggly line. This could either be caused by the "finger" being slightly misaligned (probable) or a manufacturing defect (misaligned screen, also probable and still not good).
You might say "The Eris was using the same app, why does it look better?" I would bet that it has to do with the fact that the Eris and the Droid have different screens. It is unlikely that the app has been optimized for both screens. This points to a fundamental weakness in Android - and that weakness is fragmentation. iPhone devs have the luxury of knowing the exact resolution, memory, processor, hardware, etc... when they are writing an app. When we develop for android we have to design to the lowest common denominator (usually 1.6, with lower res screens, lower CPU and mem) if we want to access the most users.
I still love my droid and silly test like this doesn't affect my experience.