Unfortunately, your question is too vague for me (and most readers, I suspect) to offer any constructive feedback. Do you mean, how many significant digits are trustworthy in path coefficient estimates? Do you mean, can the sign of the coefficient be considered as accurate? Or do you mean something else? Please elaborate.