A lot was left out of the study. I find their methodology fishy. For example, here's their test area:
Located west of the sampling site is a set of traffic lights, which results in various driving states such as cruising, braking, idling, and acceleration. Stop-and-go traffic dominates during rush hour periods, while free flowing traffic is more typical outside of these hours, especially overnight. Given the downtown location...
Downtown... stop and go for large portions of the day... various driving states... in short, even if two people are driving the exact same car in the exact same condition in the exact same driving style on average, if one at the particular moment of passing the sensor happens to be letting off the gas, while the other just happens to be accelerating when it passes the sensor, the two cars are going to give wildly different pollution readings.
I'll also note that the paper says that it's still in review, aka it hasn't passed peer-review yet.
I'm sure the general premise is right, that small numbers of vehicles cause most pollution. But I think their experimental setup is pretty bad. The stupid thing is they're collecting the data they'd need to control for it - they're taking pictures, which would let them tie vehicle plumes to particular license plate numbers, and then only study vehicles that pass by the sensor a number of times times to that they can get a running average. Another way to control for it would be to have a dozen or so sensors spaced out down the road spaced well apart so that they can average a particular vehicle's emissions on a single drive down the road. But a single sensor, single pass way to rate a vehicle's emissions as good or bad? That's a terrible approach. And they stretch very far on their conclusions based on this approach.