Superior filtering and actual maximum power point tracking.  Solar panels' maximum power point is usually about the same in widely varying light conditions.  Normally the duty cycle should be adjusted so that the voltage of the panel remains at the relatively constant maximum power point.    The approach used is to follow the maximum power point voltage using a PD filter.  This gives great dynamic response to changing loads.    The maximum power point is adjusted by 0.2 volts every 10 seconds or so and the direction reversed if the power decreases.   This seems to work reasonably well and is simple.  I do not like scanning the entire range because of the lost power during the scan, and a 0.2 adjustment near the peak power point is generally < 1% less power than the maximum.

I am achieving about 15% more power than a direct connection to the battery.

From this experiment, it seems I need to add low pass filters to the voltage and current monitoring to decrease noise before I can increase the rate of tracking.