Surely if it is a good-quality stats package, it would deliver accurate results regardless of whether it's JS, image-tag or log file-based?
If you're referring to the section headed "Get more accurate website statistics with log file analytics versus script-based software" then yes, it's a bit disingenuous. While you can say that a given server- or script-based stats package is or isn't accurate, you can't entirely compare across two such products because they're inherently not measuring the same things or even using the same source data. Interestingly, note that the text below that heading doesn't continue the "better/more accurate" argument. Obviously it's perfectly fine to compare two server or JS packages.