Hi kschuett, thanks for your question,
Your assumption is correct. The structures corresponding to the elastic tensors are the conventional cells as computed using pymatgen’s get_conventional_standard_structure method. We also recently discovered the discrepancies between the SciData repository and the online database. These discrepancies arise because the data in the paper repo was constructed from two merged databases which had slightly different approaches for constructing the conventional cell and used different versions of VASP. There are also a number of structures in the paper repo which used rhombohedral cells, which are technically not conventional. The IEEE conversion for these structures was correct, though, and we’ve implemented the IEEE algorithm very recently in pymatgen and will be providing the tensors in that format soon via the API/website.
The calculations guide is deprecated, and is referring to the data that’s stored in the paper’s data repository. Thanks for bringing this to our attention, I’ll update the calculations wiki to fix that.
We’re also in the process of working towards providing more detail about the elastic calculations (VASP parameters, structures, and fitting) on the website, which should make the starting structures and intermediate results used to construct the elastic tensor explicit.