ATK doesn't use B3LYP, so that explains the difference. Indeed, there is no experimental confirmation of the metallic nature of (6,0), or rather there wasn't in 2010 when that paper was written. To be honest I find the band structure of the (6,0) structure in Fig. 2a suspicious - I can except a small gap appears for some model and another gives metallic, but overlapping conduction and valence bands like that is unusual (it's sometimes called negative band gap) and I wonder if maybe it's a spurious result. At least it should have been explored in more detail in the paper. I respect the authors, but that figure look weird, and their argument of why tight-binding gives a different result is not convincing since LDA would capture the curvature too, and they don't explain that difference.
