As I recall, handling of DROP events in the ddl_command_end event is not completely consistent. You may even find that some DROP events are not returned by pg_event_trigger_ddl_commands. I suggest you stick to pg_event_trigger_dropped_objects() instead (for which you need sql_drop).
That's what I'm seeing as well and have a workaround in place that does just that. But to track schema changes requires two triggers and two functions. Which is fine, it works, but I'm wondering if it's a documentation issue or an outright bug.
