Yes, there is a way to influence the behavior of the null values. You can set/change these in your error configuration properties.

With that being said, it's really best practice to not have any orphaned fact records...meaning that no dimension keys should have null values. Each dimension should have a default record, and that key should be used for missing/non-existent dimension members. For the date dimension, "1900-01-01" is typically used as default.

Coming back to your specific scenario, I think it will be better addressed with a "Status" dimension or indicator. It would be fairly simple to then slice the row count by that indicator (complete/incomplete), as opposed to expecting your end users to figure it out (by means of interpreting the missing or default dates).