There is a problem in DFT that covalently-bonded species are overstabilized in GGA. This was first documented a while back for peroxides, but it has never really been dealt with systematically across other chemical spaces. It seems like the problem is also true for covalently-bonded molecular species across other anion chemistries, extending also to pernitrides, probably disulfides (like FeS2), and also perchlorides. This can result in ridiculous results on the Materials Project such as NaCl3 and NaCl7 being stable on the convex hull.
I’m not sure how to deal with this, but these overstabilities throw off a lot of convex hulls and probably mess up all sorts of stability analyses using the Materials Project infrastructure. I don’t personally have the bandwidth to fix this, but I’m hoping to bring it to everyone’s attention so that it can be addressed in the future. SCAN probably will fix this, but in the meantime I would recommend perhaps a new GGA anion-correction for these systems. There is also probably a nice publication that could result from this analysis.