Yeah, it feels like one of those immediately neat things that falls apart when you look a little closer. The data behind it is quite easy to work with though, and I think there's some better (more interactive, a bit more control) visualizations someone could come up with.

I used it myself last year when I was wondering about the emissions associated with different diets (specifically I was curious about what was more effective from a reduction standpoint: being a vegetarian at different points of time vs. owning an electric car. Since the disposition of electricity has changed pretty significantly since the early 90s in some places it varied pretty widely [the informal lit review of food-based emissions was quite a mess though]).