On the other hand, 13 minutes/K slowdown for same apparent effort when doubling the distance is a pretty typical slope to see when plotting running speed against distance on a semi-log plot. The question is whether the Ambit is actually only sampling once every 5 seconds or 1 minute, or is it sampling more frequently to get distance, then just storing distance as well as location and time at a reduced sampling frequency so as to save memory? (I will admit that if you ran down to the road intersection and turned around, and that is coincident with the one section of slower pace indicated on the graph, they are probably actually working on saving battery life by minimizing computation, rather than simply minimizing memory use).
I would expect my 2nd run to be slower, regardless of distance, since I only stopped for 30 seconds to switch modes so it was really just the end of the same run. My first 2K on any day is fast (for me) and full of hope. It all goes downhill from there. ;)
Supposedly when I select a 60 second recording interval, that is the *maximum* GPS recording interval but the Ambit may decide to record some extra points based on the movement it detects. When I'm running down a straight road and watching the distance, it seems to wait the full 60 seconds. I turned around a few meters before the T-intersection and it definitely waited 60 seconds there - and thus lost about 200 m of my route. I stopped running at our driveway but the track stopped about 175 m before that.
The 5-second recording interval works really well though!