yes, 1m for every 3.3ns error, we all know that.

the problem is those GPS vendors only provide long term average error for their products, there is upper bound guarantee at all.