It's true that the path would be shorter, but why would a 20" wide motorcycle be faster than a 20" wide car?
Transients make the difference I think; motorcycle has to shift CoG way too much in a slalom to be fast.
EDIT: Fundamentally a quasi-static lap sim is a worse approximation for a motorbike than a four-wheel car. QSLS assume that yaw dynamics are fast, which I think is not true on a motorcycle. Anybody know what order of a system motorcycle yaw is? Shifting your body to turn is something like jerk (3rd derivative) or snap (4th derivative) input I think, but not really sure.