I don't know what NASA did, but Angular Momentum can explain things a lot better than Linear Momentum.NASA speculated that the masses would have 5% of the initial momentum; but how can 5% return 100%?
At least we know: there's initially something in/with the cylinder S[cylinder,0] and in/with the sphere's S[spheres,0]. Then at time (T) there's nothing in/with the cylinder and all in/with the spheres S[spheres,T]=S[cylinder,0]+S[spheres,0]. Finally all is returned except the tether being wound the other way. This indicates (almost by definition) that energy is conserved for such situation.
As all situations/conditions are rotating around a center point, I don't see the use of Linear momentum at all as there isn't any.
You could consider a Linear-momentum-part but only if you consider it in relation to its center (because all is attached), let's say: a centripetal-momentum-part.
It needs to be as flat and linear as the surface of a sphere in relation to its center. Hence, angular-stuff.
Anyway, I'm interested if you find it closer to 3-times or 9-times.