My students regularly show me how easy it is to calculate extremely precise results. Now, whether they should trust that precision is a rather different question, and one that even the good ones struggle with for a while. (In other news, I wish that more people paid attention to uncertainty propagation. Or heck, even just some oversimplified heuristics for significant figures.)
As indicated by other quantifications in the post, I assume it is based on the average amount of time people were able to use the device before they had to stop due to sickness.
That just comes as a % related on how much (average) time people played the same game with and without the nose. 13.5 percent refers to those ~90 seconds added playtime. This means, on average the Tuscany demo has been played for 697 sec (~11 and half mins) without nose, and 792 sec (~13 mins) with.
It would be interesting to know the dynamics of the experiment (the testers were aiming to hold on as long as possible? Or they were simply told "play as much as you want"?), since the Tuscany demo is quite short, and most of people wouldn't play it for more than 10 minutes, besides sickness
