This is with the Tran-Blaha mBJ potential doi.org/10.1103/PhysRevLett.102.226401 . The problem is that this is not exchange-correlation potential but rather just an exchange one and it uses LDA for correlation and this is properly propagated to the metadata. Thus the issue is LDA gets selected in the classification. While it technically uses LDA, but it might make sense to do the clasification selection based on the highest class of the functional ladder detected?
See for example: https://nomad-lab.eu/prod/rae/gui/entry/id/X78d1GK-QJGbpM9y0sgY2w/mTCriN9mORTYAaY-Wyx0AxR-mLmW/overview
This is with the Tran-Blaha mBJ potential doi.org/10.1103/PhysRevLett.102.226401 . The problem is that this is not exchange-correlation potential but rather just an exchange one and it uses LDA for correlation and this is properly propagated to the metadata. Thus the issue is LDA gets selected in the classification. While it technically uses LDA, but it might make sense to do the clasification selection based on the highest class of the functional ladder detected?