it does not include any kind of friction losses in the engine, transmission and wheels, nor any braking losses (which turn motion energy into heat after all). Some of those are hard to model with a simple formula as they are not directly dependent on speed, hence simulation and the resulting mileage vs. speed graphs. It is true that with increasing speed, air resistance starts dominating at some point.
Exactly. The velocity cubed part doesn't actually contribute much, going by these experimental results:
![](http://www.metrompg.com/posts/photos/gcc-autobild1.gif)
(The curvature is because the y-axis plots fuel economy rather than consumption. Consumption would make the results clearer in this case).
Picking some numbers off the 535i trace at 20 and 10 mpg gives 80 and about 138 mph respectively. This is clearly very far from a cubic relationship, if fact, it's not far from linear!
Overall, I think the most important factor is engine efficiency - this is at it's best only over a small range of RPM. It's impossible to keep within that range in stop-start traffic, so taking a route that allows maintaining a steady speed in a suitable gear makes a large difference. In addition, every time the brakes are applied, momentum is lost, and money is wasted.
How much? Well, say consumption was 25% less on the steady route (it's a reasonable, probably conservative, guess). That means that the steady route could be up to 33% longer and still save money.
Thinking about a similar example where I live, to get across the city by the shortest route is about 5 miles and takes about 30 minutes in daytime, whereas the highway alternative is about 7 miles and 10 minutes. Fuel consumption is probably about the same in either case, but it's a huge difference in time!