2021-10-03, 02:04 AM
I think this paragraph will shed some light: "Output for asteroids and comets can include formal +/- 3-standard-deviation statistical orbit uncertainty quantities. There is a 99.7% chance the actual value is within given bounds. These statistical calculations assume observational data errors are random. If there are systematic biases (such as measurement timing, reduction, or star-catalog errors), results can be optimistic. Because the epoch covariance is mapped using linearized variational partial derivatives, results can also be optimistic for times far from the solution epoch, particularly for objects having close planetary encounters."