This is how VGChartz operates:
They scour the usual sites (NeoGAF, IGN, Eurogamer) for any information on sales data.
Once they have a nugget, they use their "model" to predict the sales curve.
If new data comes out somewhere, they update their curve point.
VGChartz is as reliable as googling a good source for an available information.
If there's no good source (some real chart tracker or PR) they are in the dark and their stuff is just guess.
During the whole Virag thing, their data put GT6 at 3.X millions. So no, their data wasn't "almost spot-on".