I think I know what the OP is talking about . The part number he quoted is the part number engraved on the top of the crystal on an official arduino UNO. Apparently he has a stopwatch sketch running (or plans to run one) and wants to know the accuracy. I found this post with a Google search of that part number:
http://forums.adafruit.com/viewtopic.php?f=25&p=121801
I assume the oscillator is some division of the clock crystal, which says SPK16.000G on the top - presumably 16 Mhz. The hardware says Arduino UNO, board model R2.
Using the above method above I can get +/- 1 second every four days. Maybe better if I kept refining the TIME_CORRECTION value. But I think JChristensen is right - without temperature compensation (e.g. Chronodot) that's about the best I can do.
I agree Westfw - an amazingly small error is unacceptable in a clock application! A mechanical watch that's banged around and encounters large temperature variations keeps time as well as this.