I don't have access to Indinero's data, but for my own startup, Compete's data is no more accurate than a random number generator. Both the absolute numbers and the shape of the graph have essentially zero resemblance to the real data.
That being said comparing compete data to hundreds of parked domains that get enough traffic to have a decent data sample yes I would agree the data is off. But not anywhere near enough to make any difference in the point I was making.
I'm pretty sure for a "decent sample" - at least, accurate enough for the data to be useful for anything - you need to be at least a top 100 site, and even then I've heard stuff from a couple webmasters of those sites that Compete data is off by an order of magnitude or so.