There are two situations here. One, is that the rider is not pedaling and just dropping of the curb or bunny hopping. In this case there is no, or very little, movement in the pedals so although there’s pressure on the axle there is no work being done. This is why the cadence sensor in each pedal is needed. The second case is that you pedal off the curb (not really applicable to bunny hopping). Here the pedaling action is more removed from the action of dropping of the curb. Power will be calculated normally and the drop will be considered a “vibration.”
The unit needs to filter out vibrations if it’s going to measure the specific type of work and power we’ve come to expect our power meters to measure. That is to say to measure the work and power done to move the bike forward not to balance or absorb shock and vibration. To answer your question, data will be “cleaned” and unnaturally high spikes are calculated out, smoothed or disregarded.
This all touches on the reasons we really need to wait and see some third party real world testing and data before we draw any conclusions on overall accuracy.