I don't know what compass you were using, but maybe you simply had a too refined training setup? For example, the teacher could first measure the apparent Azimuth of the target from the start location with that exact compass and give you that number along with that very compass, even though the true azimuth of the target was off by 5 degrees, e.g. because of the compass dial excentricity. (by the way, the mirror already does not belong to the classic model we discussing, as it eliminates several obvious issues to some degree). Even a 0.1 deg. accuracy is achievable on a level table (with big enough dial), especially if you care to account for your particular compass pivot drag. But when you are on a trail and taking bearings every so often on the go, you will see my numbers or even worse ones. And the price of the compass wouldn't help to improve that if it is of a classic type with less than 2" needle. I believe that being conservative on your compass capabilities is better than rely on its ideal test conditions technical specs blindly.