Well from my side what I am thinking on flat is also like you wrote in your point #2. Even with two good devices there is always a notable difference when running at slow speed. So I always wonder how much arm movements can affect the total distance. Cause, if I am right, the proportions of arm variation in total distance should logically decrease with higher speed (ex if gps record every 1sec.) and this is what I see in my tests. As for your watch 615 is probably as good and reliable as 625, but the behavior I am talking about is more pronounced on 505.