Data searching in the primary literature and authoritative databases (such as the CRC, DIPPR, NIST) is messy enough in its own right. Going out onto the Web and using the plethora of free chemistry databases and search engines is even messier, and can lead you in circles.
Chemistry databases such as ChemSpider, ChemBioFinder, PubChem, any number of MSDS and supplier catalog sites, plus crowd-sourced tools such as Wikipedia, are very tempting. Google searches often bring up content from these sites. But data found in them are typically copied from other sources uncritically and are often unattributed, and therefore suspect. Inaccurate data are easily grabbed and replicated by automated web crawlers and inattentive authors and editors, and the Web can then magnify those mistakes many times over. If you can't trace a value back to some kind of authoritative literature source, you should not trust it.* (MSDS in particular are unreliable in almost every respect.)
Such errors not unique to the open web -- peer-reviewed literature is not foolproof either. Science is a human endeavor, and humans make mistakes in both measurements and assumptions, not to mention typos.
Moral of the story: Trust nothing implicitly. If the data looks wrong, it probably is.
* To see an illustration of this point, look at this blog post from a professor teaching this concept in a chemical information class.