This is probably because it used the old incorrect method to calculate the kinetic energy from four years ago in BFit.
The new-updated version was added two years in BFit with tests matching the kinetic energy value obtained from the slater files.
Hope that helps
It appears there is an issue with how the kinetic energy density (KED) is being evaluated in the Slater dataset compilation script.
This problem shows when comparing radial grid plots of the KED from the Slater set to those generated from the Gaussian or HCI datasets:
This is the function that needs revision: https://github.com/theochem/AtomDB/blob/da1a133cfaf7a60767a565ade9fa26b31ed93e40/atomdb/datasets/slater/run.py#L552