rather than blanking them out entirely.
This lets csv-to-json.py handle the case where an earlier input file has (e.g.)
a registration value which does not match the computed value, and then a later
input file has a value that does match. In this case we want to override the
older value with the newer value, then notice that the registration can be
omitted when writing the database. Previously in this case the older (incorrect)
value would be used.