The injection accuracy of the launch vehicle is typically measured in m/s, as the velocity change required to get the spacecraft exactly on the desired trajectory. This wraps up all the dimensions of the error (the energy, the trajectory plane, your time of arrival, etc.) into one number.
From my recollection, that number is on the order of 1 to 10 m/s. Pretty damned good. The larger end is for solid-motor upper stages that you don't get to cut off when you like. (E.g. Delta II.) You should be able to find this number in the payload planners guide for the respective vehicle.
As for the propagation of the error, just add 1 m/s in random directions to a trajectory and see how far off it gets by the end. The $\Delta V$ to get back on track grows quite a bit over time, so the first trajectory correction maneuver is done about ten days after launch to limit that growth, but still provide some time to become acquainted with the spacecraft before jerking it around.
As for tracking uncertainty, this is represented as an ellipse on the B-plane of the target (e.g. Mars), and the uncertainty in the time of flight, i.e. the exact second the B-plane would be crossed. (Note that the B-plane numbers assume that the target is not there gravitationally.) For the MER maneuvers at launch plus ten days, the orbit determination uncertainties were on the order of 3500 x 1300 km, with a time-of-flight uncertainty of about 1000 seconds. Those are all $3\sigma$. I don't recall when the data cutoff was, but those would have been the result of about eight days of tracking.
We know where the center of Mars is to about 100 m, due to tracking many orbiters there over a long period of time. The Moon we know far better due to laser tracking of retroreflectors left there by Apollo. I don't know the number, but better than a meter.
