Per comments from MK, I've generated a new dataset excluding one name access point (BF, a Barkerville staff person); asked MK if there are other names we should exclude. I've also managed to pull in a bit more data on creators and condition of the object, and I've used the accession number to generate the identifier instead of the unique database id.

As expected, I've found a couple of non-unique accession numbers in our set. These items:

HER33797 (1982.0273.0001. A-B)
HER33798 (1982.0273.0001.A-B)

have accession numbers which differ only in the presence of a space, which we don't want to include in our identifier. Similarly:

HER49298 (1993.0002.0004. A-B)
HER49299 (1993.0002.0004.A-B)

I don't want identifiers that are distinguished only by (for instance) the presence of an underscore, so I've added "_1" and "_2" to the end of these to distinguish them. Waiting for more feedback from MK on this before attempting to import into AtoM.