You can start a lot of arguments on the subject! I suggest you google "Time Nuts" and look through some of the contributor's pages. In my experience, the "experts" seem to care mostly about stability, not accuracy, for their argument is that they can calibrate out the inaccuracy (once they measure it). A good OCXO usually has nice short term stability, but the Allan Variance (a measure of stability) tends to climb back up at longer time intervals. The GPS disciplined OCXO tends to force the Allan Variance back down.
I also have an HP 53310A and consider it one of HP's most "under-advertised" pieces. I also have its cousin, the HP 5371A, which can do a few more things, but runs a lot slower than the 53310A.
A bit of caution on a used one, however. The Boeschert power supplies are notorious for leaking their electrolytic caps. I've had the occasion to work on two 53310As, and both supplies were toast. Fortunately, they also can be found in a lot of other HP pieces, like some of their scopes and logic analyzers.