So sigu (used by calcem) is acting up, like not producing accurate results and with every of the many many loops it requires, the errors keep on compounding. In order to get my version of sigu to match NJOY's, maybe consider going from double to float (?) but not sure how ultimately important this piece is.