I can answer this. Each of these markers comes into the software as part of an unconnected point cloud, which then needs to be identified and 'solved' to a constraint skeleton to get the true angles of rotation and the 'stick person' representation that you typically see.
Because of this, if you placed markers with bi-lateral symmetry, it would be much more difficult to solve joint rotation angles, as the overall facing of the point cloud would be undetermined. So, you have additional markers placed on none moving (not a joint) locations that ruin symmetry and allow facing to be easily solved. Typical locations are outside of the thighs, upper arms, and occasionally the back.
I read your answer before commenting and didn't think you were so much wrong, but dangerously close to right, just off slightly. Interesting enough, each studio and sometimes even each operator, will have their own preferred marker placements. We will also vary these markers when doing multi actor scenes, so to help determine the tracking of the individual actors as well.
Marker placement in motion capture is often intentionally bilaterally asymmetric because it makes automatic labeling of markers easier (with respect to side/orientation of the subject).
Source: Am a biomechanist with several years of mocap experience.
Look lower down at the pictures or pitchers (heh). The legs themselves aren't symmetrical.
So they probably tweaked the points so that the most common range of angles to make the math easier to do.
Like I said, just a guess. But with one leg bent and the other stretched out, they probably form a line close to parallel with the ground for right-handed pitchers at the point in time when the pitcher is transitioning from wind up to release.
