So we're left in a position where there's a difference between the prediction and the measurement, when the measurement is highly accurate. That means something is out, and that might be extremely important for long range flights.
I'm not too sure how to work some of these things out, but it's possible that a few mm/second difference really adds up when you're on an enormous trip. If you're out by 10mm/second here, then your next flyby will leave you even further out, then the next, etc. until you crash into a planet. Cassini was out by under 1 mm/s, but its flight path would have been 2 billion miles. These things travel for years, and can't make too many corrections.
More info on the crazy journey it took: http://www.theplanetstoday.com/cassini_flight_path.html