The normal approximation

So I ran into the problem of proper normal approximation. Since it is highly advised to stay away from generalized implementations of the normal distribution in some softwares, I decided to implement an algorithm by myself to continue with the analysis of volatility and spot prices data. Why am I stuck? I used the Hart algorithm (which is extremelly precise to all the decimal places) that I found in Haug book. Not giving much thought that the algo will not run properly, I ran into a negative option value. And how is this possible?

Well, certanly there were several studies done on the mentioned topic and I then I decided to copy paste the algorithm from a different author (West) and came to find out that I obtain the same wrong number as I have with the Haug version. I still do not know where is the problem, but having negative values extremely disturbs me because then the rest of calculations will definitelly be wrong. Anyhow, I wanted to share a paper that was a critique on normal approximations and an analysis of different approaches.

http://www.codeplanet.eu/files/download/accuratecumnorm.pdf