On weak epsilon nets and the Radon number