This video appeared on my Facebook feed today, and as I was a bit bored at work I did the maths and it turns out the 29er was about 2% (or less) faster than the 26 MTB. Also, we get an idea of the scatter in the measurements of the 3 wheel sizes in this study from the fact that the 650b was 19 seconds slower than the 29er (even slower than the 26"), which suggests a 1-sigma uncertainty of about 2%.
https://www.youtube.com/watch?v=kxfrykeSNCEAnother, probably better study in Switzerland (more riders) found 29ers are just a few % faster than 26" MTBs, which is consistent with the above.
What's curious is a lot of people swear that 29ers (and to some extend 650b) are vastly faster than 26", by 10% or more. Why could this be? Perhaps the larger wheels truly shine over longer distances than used in the tests, which are probably not representative of normal riding in that they use very short (a few km) courses.
I'd speculate that the larger wheels might not be much faster per se, but conserve more of the rider's energy and stave off fatigue for longer when riding over longer distances. Obviously, doing a 3 km loop isn't necessarily going to detect effects such as these (assuming they exist).