This solution also does not meet the requirement. Applying filters to the already-imported worksheet rows in Excel (or attempting to reduce rows in the workbook UI) does not change how Power Query initially samples and infers the JSON schema during the Get & Transform steps. Power Query's schema detection typically happens on a preview/sample of the source data when the query is first evaluated. If the JSON property you need does not appear in that sampled subset, Power Query will not create a column for it even if you later filter the data to include rows that do contain that property. In short, post-export filtering in Excel is not a reliable way to force Power Query to detect and expand missing JSON properties.
The reliable approaches are to increase the number of rows Power Query uses for schema detection (Query Options / Data Load / preview/sample settings) or to explicitly parse each AuditData cell with JSON.
Document (or equivalent M functions) and then expand the record fields-these approaches ensure the specific JSON properties are surfaced as columns. Reducing rows via Excel filters does not address the sampling/schema-detection behavior and therefore fails to satisfy the requirement.