i think this generally will be okay for research grade observations, but if you’re dealing with some atypical RG observations or with observations not yet at research grade, you’ll have to keep in mind that num_identification_agreements works using the community taxon as the basis for agreement (see https://forum.inaturalist.org/t/num-identification-agreements-field/9457).
if you use the API, you can get a little more granular in determining what kinds of agreement you have (ex. https://forum.inaturalist.org/t/how-to-exclude-observations-with-disagreeing-ids/15016/12), but that determination also can be done only after download.