On Multiple Solutions in the Determination of Orbits from three Observations