Why do you think that truncatingStirling's series is preferable to approximating the more accurate Lanczos asymptotic series (which is valid for Re(n) > 0)? It is more 'robust' to calculate the logarithm of the gamma function because you are less prone to overflow for large values of n.