This VGChartz article bugged me as an example of bad and sometimes contradictory advice.
Lesson 1 - Don't Take Six Years to Make Your Sequel
Gran Turismo 4: 89.61%
Gran Turismo 5: 84.69%
First of all, a MetaCritic GameRankings score (why is he using GameRankings instead of Metacritic, anyway?) drop of 5% is not necessarily statistically significant, especially since there's no way, from just the scores, of verifying that the same selection of outlets or reviewers were included. The drop may indicate a drop in popularity of the genre, or many factors other than the development cycle.
If what he means to say is that the earlier you release, the earlier you get paid, then he's right, but for every GT5 that fails to meet expectations or Daiktana that becomes the butt of jokes or DNF that fails to materialize, there are games rushed out the door lacking polish or originality. Of course, since the subject is sequels perhaps originality is not a factor.
Sure, six years might be extreme, and there might very well be a point of diminishing returns, where more development time doesn't necessarily improve things. However there may very well be cases where 2 years is not enough, and where even three or four may be justified. Might not another 6-12 months have improved, say, Halo 2?