Interest has grown immensely for the application of metal organic frameworks (MOFs) in numerous gas adsorption and separation processes, such as CO2 capture. High throughput computational screening of MOFs for these processes is now commonplace. For this purpose, several databases of experimentally characterised MOFs have been developed that contain tens of thousands of structures that have been ‘cleaned’ for computational screening. The potential material pool has expanded even further towards the range of millions of structures with the introduction of hypothetical MOF (hMOF) databases. Nearly all hMOF structures are generated using combinations of structural building units (SBUs) extracted from experimental MOFs, which have been found to contain high rates of structural errors (i.e. excluded protons, disordered atoms, improperly labelled framework charges, etc.). Undetected errors pose a considerable threat to the accuracy of results obtained by previous— and future— MOF screenings as they completely alter the system being studied during periodic DFT calculations (i.e. different number of electrons, nuclei, etc.). This work introduces a manual investigation effort aiming to identify common errors present in hMOF structures and study how these errors influence electrostatic potential fitted charges and GCMC-simulated gas adsorption properties. A novel repair protocol was developed to generate a library of over 1500 'HEALED’ SBUs that are ideal for the generation of chemical accurate hypothetical materials. This library represents one of the largest collections of SBUs in the field to date, and it represents an important step towards improving the reliability of HT hMOF materials screenings.