Given the manufacturers are building in situ into the software, they likely have similar data to what you found. This seems to suggests that an audiogram isn’t directly useful for programming aids which is consistent with the need for REM. If in situ can approximate REM, or even surpass it, then REM and possibly audiologists become less necessary and both DIY and OTC could expand bringing costs down. It sounds like the primary challenge with in situ is the subjectivity of the person being fitted.
Were there any clear points of difference between REM and in situ or was it all within the margins of error?