Best-fit method; vector or least squares?

[dbusch]

I need to understand the difference between these two methods. I need a 3-D best-fit alignment on a compound-curved surface that will tell me about the form only (position not needed). I have 163 vector points, and the results look excellent with the vector method, but very bad with least squares. I read the help page on it, but I'm wondering if the vector method is not showing the error vectors where the nominals are? Are those error vectors the T-values? Are the least-squares error vectors comparisons to one average error value for the whole surface? Which method gives the most accurate form measurement? Am I heading for insanity?? Any insight very much appreciated!