The reason for this is that statistics can easily be manipulated to tell a story that a particular source wants them to tell. Therefore, you need to have evidence from a source that is not trying to prove a particular point.
To use a hypothetical example, let us imagine you are looking at the rate at which students drop out of school. A source that wants to dramatize the problem might look at a cohort of kindergartners in a given school district and then ask how many of those students graduated from high school in the same district thirteen years later. A source that wants to play down the situation might look only at students who started high school in a given school. They would then not count those who transferred to another school. The first method would yield a much greater dropout rate than the second.
If you do not use evidence from a disinterested source, you run the risk of getting biased data.