Hi Martin,
Thank you for taking the trouble to explain. I am happy with this because a more accurate 1 minute simulation of 3D schadows probably wouldn't effect the hourly, dayly, monthly, yearly or lifetime results much. However allow me to suggest an alternative future improvement to get rid of these "hourly discontinuities": instead of using hourly shade values you could first fit a curve through the hourly values to generate "continuous minute values" during the main calculation.