Steve Wexler has just published a post titled “There is no perfect chart and there is no perfect dashboard.” Go ahead and read it.
Back already? I agree with with Steve, but I'd like to stress, as he suggests, that it's possible to sort visualization types according to how effective they are at conveying a specific message, or at enabling certain tasks. The decision of how to transform data into spatial properties (that's what visualization consists of) is largely subjective, but it can't be based just on personal aesthetic preferences. There is some science behind it, no matter how imperfect some of it may be. Context and purposes also matter.
Let me give you an example, this visualization by Anna Vital (don't miss her posters!) for Google Trends*. It's an interesting, attractive, and elegant graphic, but I think that a few changes could make it even better.
See the screenshot below. The fact that the lines are shaped as curves makes 3h 9m —close to the center of the circle— look only slightly longer than 1h 27min; and 1h 27min looks shorter than 51min, just because the latter is in the outmost ring. Besides, the little clock at the bottom may be a misleading cue, as these lines don't depict times of the day, but time after each of the attacks happened. I think that in this case some of the visual appeal of the graphic ought to be sacrificed in order to increase clarity: lines could be straightened out, or there could be a circle per attack, so proportions would be respected.
I discussed some of these issues (and many more, related to other visualizations) in this recent conversation with the Google News Lab folks.
*Full disclosure: I'm a consultant for the Google News Lab, working mainly with Simon Rogers.