Are you referring to the Registry data set in the Endpoint data model?
If your data is incorrect, you have field aliases, calculated fields, or field extractions on tag=endpoint tag=registry events that produce incorrect results. This could include automatic extractions on raw event text like "user=foo.exe."
Start by determining which field-value pairs identify tag=registry. You can use Settings > Tags > List by tag name in the UI or run 'splunk cmd btool tags list --debug' on your instance. From there, review and correct knowledge objects for the related events.
You can rebuild the data model or leave old, incorrect data in place while new events populate the data model correctly going forward, depending on your requirements.