Count me among those who have been under the misconception that ball bearings accelerate quicker, have less constant speed friction loss, and decelerate more slowly, compared to journal bearings.
The better transient response is a result of the BB slowing less rapidly, leaving it at a higher rpm after a given time period, and accelerating more rapidly, which when combined with the higher end state rpm results in a desired higher rpm being reached more rapidly.
The percent improvement Garrett quotes for acceleration, 15%, is inline with what I have seen in other studies that compare BB with JB. The amount of time difference may not noticeable to the driver, but the measured difference is more than a 'very small difference'.