You should adjust your belief to fit the available data, don't adjust the data to fit your belief.
I would normally agree, but in my opinion, surveys fall into the "Lies, damned lies, and statistics" category. They hold the same value as apache-bench numbers that are posted every time a new webstack appears.