This is a use case brought up by @ltalirz.

For CifData, which doesn't store data on an atomic level, it's sometimes preferrable to not store atomic site information and more, since the represented structure hosts up towards 1000s of atoms, which would result in suddenly adding a huge amount of unnecessary bytes to the database and also make the search slow.

On another note, I think this may be possible by using aiida-optimade calc, where individual OPTIMADE fields can be specified.
However, since some fields rely on others, field X might still be calculated even if not specified due to Y's dependence on X. (E.g., the list of cartesian_site_positions MUST be the same length as species_on_sites.)