I mean, how is this true if he doesn't know which watch is "wronger" and by what margin of error? If it is 2pm and one watch shows 2pm and another 1:30pm the average is worse off than having one watch that shows 1:46pm.
If you have 2 watches with let's say 5-minute difference, I have a pretty high confidence that the indicated hour is correct. The chance that both watches e.g. broke in the same way is small.
I don't think you will necessarily get a higher precision, but your confidence of the indicated time is higher.