In general, this is a draw. With certain rare exceptions, we can't state an obvious advantage of either mode. Moreover, there are no trends or regularities related to application class, including single- and multi-threaded nature, which could at least conform to the official positioning of these modes. The 0.1% total loss of the Ganged mode doesn't indicate anything, since another application can turn it into a complete draw or even a victory.
Thus, our conclusion is brief: in terms of real-life desktop software, it's generally the same, in which mode Phenom's memory controller is working. There are certain rare exceptions, but their character can hardly be predicted (at least, based on our test).
As for the question if we were right, the 0.1 point (0.1%) difference is hardly a reason to start using the Unganged mode in primary tests. Especially as the Ganged mode is motherboard default, and even the 10 times higher value (1%) would still be negligible as a possible measurement error.