Another thing to point out: my heart rate was about 10 beats higher on the faster ride where I supposedly produced less power - and Strava had that data the activity pages. Based on speed and heart rate but not wind, I have no idea how they got that figure. Well, I do: either they plucked it out of thin air, or they changed their algorithm between the two rides.
[ EDIT: Actually, thinking about it, do know what caused the difference: the first (slower) ride had a stronger wind blowing up the course, so I went slower in the first half and faster in the second - as you can see if you follow the link. Strava say (your link) that they assume no wind when calculating power, so my slow speed in the first half has dropped my estimated power, but the cubic speed/P(aero) relationship has given me a massively inflated power reading in the second half: Wiggo's hour power to win the Tour de France is less than I supposedly put out in the second hour of my ride. ]