I've lifted both of the graphics you'll see below and the core of the argument from this Ilan Goldenberg post, so all credit for the work is due to him, I just don't think he made the point as clearly as he should have. To get to the heart of the matter, just take a look at this map that General Petraeus offered as part of his presentation:
That map shows Baghdad awash in sectarian violence in December of 2006, and it shows the violence steadily decline over time until August of 2007, where it's still certainly a problem but a much reduced one from where it had been before. But notice something funny about the map . . . the color-coding of the neighborhoods as Sunni, Shiite, or mixed stays constant throughout the period even though it's a period during which we know there was a lot of violence and a lot of internal displacement. What would happen if we showed how the neighborhoods changed over time? Fortunately for us, General Jones prepared maps that did just that for his own presentation:
Jones' maps show the exact same downward trend in violence as Petraeus' do. But they also show something else. In particular, they show the disappearance, over time, of mixed neighborhoods with violence, refugee flows, and ethnic cleansing producing a city that's much more starkly segregated along sectarian lines than it was twelve months ago. In short, the number of incidents is plausibly declining not because of improved security, but simply because there's relatively little fuel left for the fire. Note in particular that Petraeus shows a large decline in violence between December 2006 and February 2007 which is too soon for the arrival of the surge forces to have made a big difference, but which coincides with the disappearance (shown on the Jones maps) of most of the mixed areas east of the river.