You aren't doing anything wrong logically, but Databricks requires row-level changes to happen on Delta Lake tables. Standard Spark tables (like those backed by raw Parquet) are often immutable. Have a look at the Language Manual for DELETE to ensure...
