Calculations of Luc Arnold, confirmed by Toshimi Taki and Richard Schwartz do indicate that the residual error for the 18 point support should be 20% higher. This was found out by using a finer grid in the FEM-software, so acquiring more accurate results. Up to now the other results have not been recalculated, so care is taken not to generalize it, only one particular case was calculated. We hope Toshimi will further refine all results in the future.
Luc Arnold was using support pads of 10 mm diameter on a diameter 680mm mirror, Richard Schwartz has also found out in a simple case that by using support pad of half the thickness of the disk one can obtain a advantage of 30% Peak-to-valley in that particular case. So it seems indeed to be better using not a sharp point tolay the mirror on but a ring. Luc Arnold is also indicating the support position tolerance being in the order of +/- 1 mm in order to have an error below lamda/20 surface !!!