It seems that your AV is much smaller than EV, and there is error in the formula for AV calculation. The formula should incorporate this condition as" =IF(((B47*Y64)^2-(B50^2)/(E43*G43))<0, 0,SQRT((B47*Y64)^2-(B50^2)/(E43*G43)))."
After revision to the spreadsheet ,your study TV is still too small(0.0009) when compared to the Spec TV(0.005); the GRR is still high. If use spec TV as computation basis, the PV must be recalculated(=sqrt(spec TV^2-GRR^2));both the GRR% (5.91%)and ndc(23) results come out pretty good.
However, your R data indicate that the EV is not stable, part #5 R(0.0020) is beyond the UCL(0.00129), and the resoluation is not sufficient (only 3 different values:0.0000, 0.0010, 0.0020), you might need to investigate such situations.