Yes, please add the seconds to the “observed on” time! This shall also apply to the downloaded csv file.
I wrote a code to update the metadata of my offline picture archive with the iNat identifications and related taxonomy.
I created an excel list of my picture archive that I match with the observation file downloaded from iNat using the observation time (“date taken” for the offline file) and
feed it to the ExifTool to update files on disk with the iNat data.
It works great but because the time only goes to the “minute” I have several duplicate matchings (more than I would have thougth, almost 5% of all observations), that I have to “discard” (or process manually).
The only other field in the iNat download that can be used to uniquely identify back the file in the offiline archive could be the tag_list but that works only until the offline and the online tag list doesn’t change (that happens already after the first update of the offline archive with the iNat metadata).
The issue was also discussed here: https://forum.inaturalist.org/t/is-it-possible-to-retrieve-the-original-filename-of-an-observation/4419