A variety of factors can interfere with analytics-style tracking, so analytics stats should never be expected or treated as a precision-headcount metric, but rather just as a basis for relative comparisons against each other. The confounding factors will be relatively consistent across a large enough number of visitors, making for a fairly consistent "margin of error" in the analytics stats that you'd compare, where you're really only looking for huge, glaring differences -- see this help page for a more detailed discussion:
http://www.e-junkie.com/ej/faq.tracking-how-why.htm
