Personally, I'm more inclined to think that dangerous drivers tend to speed more than safe drivers, hence the correlation, but in any case I'd love to see some evidence either way.
Urban speeding does increase accidents because of the large number of unexpected events in an urban setting (pedestrians, cars driving into intersections, etc).
But the main issue I have with speed limit studies is that seem to make this underlying assumption that everyone follows them. If the core group of accident-causing people drive fast/drunk/etc regardless of the rules, then lowering limits/changing rules must have limited effect.
Underlying all this is the government love of speeding taxes. The only problem is, the marginal return on speeding tax decreases as the tax take goes up. Because people decide to obey the limit, the tax take starts to go down. Thus the conundrum for governments - do they set the fines at a level where people are annoyed but don't mind, or do they increase the fines to the point where revenue steeply drops.
I definitely agree that the worst 20% of drivers create 80% of accidents (made up numbers), which makes most any speed limit conversation null and void.
I'd like to add another possibility beyond what you've noted. Might it be that speed isn't itself more dangerous in that it "causes" accidents, but when an accident occurs and higher velocity is involved, that means more kinetic energy is to be dissipated, and with all that energy in play, the results may be more dire?
I'd also like to see more variables factored out of the data. I'd like to know about someone (like me) who observes other safety rules like seatbelts, proper lane discipline including signaling, safe following distances, etc.
It seems to me that faster speeds may also force people to pay more attention, as each driving decision occurs in a shorter timespan.
As for urban/suburban speeding, the risk factors (pedestrians, children, intersections) are completely different, and increased speeds would clearly increase risk, as well as slow overall traffic flow as fewer cars could successfully turn onto fast-moving streets.
I can think of other situations, for example highway driving, where going 80mph in a 65mph zone probably has an only minor effect compared to going 65mph, and depending on traffic could be safer.