Choose columns vs remove columns

Best practise is to choose the columns that you want to keep, and don't make any reference to unwanted columns. That way if the data source were to change for one of those unwanted columns (deleting or renaming), it won't break your model - nothing more annoying than having a refresh break because of a source change for a column you weren't even interested in. Moreover, this also protects your model from the appearance of new (and unwanted) columns in your data source. If you're getting down to your preferred columns by explicitly removing the unwanted ones, then if a new unwanted column is added, it flows out into your data model until you update the code to add it to the remove list.