The motion capture shown is kind of limited. The wheels are tracked, but not the CG of the rider. Also, the video doesn't show the beginning of the ollie. It looks like, in this one, the rider did this entirely by loading up the springs in the trucks to get a bounce, rather than dragging the tail of the board.
It's a good physical simulation problem. Now that Gazebo has a real physics engine (by Mike Sherman from Stanford) rather than ones from video games, it should be possible to do an ollie in a robot simulation. (The game-type physics engines use an impulse/constraint model for contacts, which is unrealistic when you have a collision followed by sliding.)
There are bushings which will provide a small amount of spring, but your general points stands that the force comes from pushing down on the tail and shifting the front foot up to the nose to pull the skateboard up.
No, it doesn't make sense to track the center of mass of the skateboarder here. If you're interested specifically in the forces on skateboard, you want to track the center of mass on the skateboard. If you include the skateboarder in your system and track the new modified center of mass, you will only see the effect of gravity, being the only remaining external force.
Tracking the centre of mass of the skateboarder would give you an idea of how to get additional height on your ollie. Essentially sucking up your legs, bending at the waist and throwing your arms down will allow you to clear higher obstacles.
It's a good physical simulation problem. Now that Gazebo has a real physics engine (by Mike Sherman from Stanford) rather than ones from video games, it should be possible to do an ollie in a robot simulation. (The game-type physics engines use an impulse/constraint model for contacts, which is unrealistic when you have a collision followed by sliding.)