So this is the core problem with all of the general relativity materials that model it as a rubber sheet causing curvature in spacetime. They always model it with focus on _spatial_ curvature: which is totally able to model an orbit or a hyperbolic trajectory as a geodesic, but it totally cannot model "throwing a ball up" since the geodesic for throwing a ball up is just a straight line.
The important thing is that gravitation is a distortion in space-_time_, which is way trickier to model as a rubber sheet because you end up with one dimension of space and one of time. If you distort _those_ (also, they don't distort quite like a ball-in-a-rubber-sheet), you can get the results of a ball being thrown up. It's also possible to visualize this for 2 spatial dimensions with a distorted 3d space, but tricky.
The important thing is that gravitation is a distortion in space-_time_, which is way trickier to model as a rubber sheet because you end up with one dimension of space and one of time. If you distort _those_ (also, they don't distort quite like a ball-in-a-rubber-sheet), you can get the results of a ball being thrown up. It's also possible to visualize this for 2 spatial dimensions with a distorted 3d space, but tricky.