In many cases, some columns are selected or generated during joins just for filtering, but shouldn't be included in the output. A common example is A not B type joins, where I need to left-join, then filter where B.key is null. B.key should be excluded in the output dataset, but has to be included in the select list in order to perform the filter. By adjusting the workflow of the join recipe, these common types of joins can be set up without needing a prepare recipe afterward to reduce the select list.
