That figure is extrapolated based on the longest photon path length through the device. It isn't based on real experimental results (the device was demonstrated with n = 4), and it does not take into account the exponential quantity of photons you'd need to pump through the device before getting a measurable result.

Fundamentally, there's nothing this device is doing that a classical computer couldn't, so there's no reason the exponential base should be any different under physically realizable conditions.