Hi Xantix,
If you really want to do this, you can turn on template access by going to the Dimensions in the Library, select the Lookup you are useing in your fact build, open the property sheet, and select the Data Access tab. There is a checkbox that lets you use template access.
However, I suspect that Include Unmatched Members is not really the solution you are looking for. That will essentially elimnate all checking on your lookup. "Anything that comes through is OK with me!"
From your description, I'm guessing the field that is sometimes null is the "business key" of the dimension is that correct? The reason I suspect that is that the business key cannot be null. Generally, any kind of lookup will fail when it attempts to find a null value business key.
Your problem is similar to a missing date where that is legal. For example, you may always require an OrderDate, but ShippedDate may be null until the order is shipped.
So a lookup on a null ShipDate will fail but you still need to indicate that it is missing and you still want to be able to analyze on that information.
In such cases, you usually want to replace the null value with some known, fixed value like zero or "?". You can do this in your source data by placing a calculated column in the SQL or by using a derivation in a pre-pass fact build (since you can't use derivations as a dimension element in a DS 6.5).
You will also need to include your null-substitution value (e.g. "0"

in your dimension. You can do that by adding a static member at the lowest member of your hierarchy that you use to load the dimension table.
Hope I'm on the right track and that it helps.
Matt