Here's a thought.
You measured them after sundown and under load, then again a while later, also under load.
If there was some gas, left over from the charging, trapped on the surface of the plates, it would produce a raised resistance and a voltage drop from the load current. With time under discharging conditions the gas would dissipate and the resistive vlotage drop would be reduced. If that change in voltage drop is greater than the drop in voltage from the discharging, the result would be an observed raise in voltage at the terminals. The actual cell voltage would be dropping, but the resistive voltage drop would be overcompensating for it.
Another possibility is that the cells are warming a bit just after sundown (perhaps from resistive heating), which would actually raise the cell voltage slightly.