Recently I received a request from a reader related to the handling of missing or invalid dates in Tabular (but the same can be applied to Power Pivot). In the Analysis Services Tabular book, we wrote that you can insert a special row in the Date table with an ID that can be used to map all the rows in the fact table that don’t have a valid date. Unfortunately, using NULL in the Date column is not compatible with the “Mark as Date Table” feature of the data model, which is necessary for Time Intelligence functions in DAX. The alternative is between using a special date (such as January 1, 1900) or removing the row at all from the Tabular / Power Pivot data model.

I described this last approach in my article Handling Wrong or Missing Dates in Tabular on SQLBI web site.
What is your favorite approach? Comments are welcome!