Sounds fair, but that being said I do not think it restricts us from creating a DLT code as a SQL task in a Job. Once the job is evaluated/dryrun/run (which has a SQL task with DLT definition) , I immediately see an automatically created pipeline in the UI. (Streaming tables created in Databricks SQL have a type of MV/ST.) 

So don't I get the benefits you outlined in this case too but without the efforts of creating a ETL pipeline ? Which are the benefits I miss?