I have noticed, on the occasions that auto stop has failed to activate, that the indicated trip mpg can visibly drop as fuel gets used and the increase in miles is very small. And that's after, say, 50 miles of open road motoring with a reported mpg of around 60. So the auto stop can give some benefit to mpg in real life particularly if in a long queue for traffic lights where there are significant stop periods.
However, traffic snarl-ups are often in the form of slowly crawling traffic, such as the queue for a roundabout, when the auto stop doesn't get a chance to work. I wonder whether the proposed improved emissions testing will include some of this.