As said, minor changes in velocity (on paper, or in Chairgun) cannot account for those large variations. And my experience has been very similar: rapid vertical shifts e.g from centre to top edge of kill at modest ranges.
The more I read this thread the more I'm convinced of the following:
1. Springers are complex systems, verging on chaotic. Some might say capricious and go on to suggest the work of diabolic forces.
Chaotic in the sense that tiny variations in input can lead to big variations in output.
2. Most of the odd behaviour reported is I believe not attributable to one single cause. People try to logically work out the cause by removing variables until they get to changing a single factor at a time, which is sound methodology. But even then they often can't find a correlation.
I think the changes in velocity result in the pellet exiting at a different point in the rifle's movement. Muzzle velocity doesn't give the whole story here, as two identical springer actions might give identical MV but one action moves faster than the other. The fastest isn't necessarily the optimum of course. If during recoil (forward piston motion) the muzzle is rising, and then during surge (rearward piston bounce) the muzzle falls, you want the pellet ideally to exit when it is pointing in the same place as when you break the trigger. This could be why minor temperature changes can have a big effect. You also want to minimise the amount of vertical muzzle movement which you can do by balancing (adding muzzle weights) or adding mass to the whole system.