There is the rare occasion that
(ok, we have been faced with that archaeo-technical, legacy-system-related issue already in some of our very first SSAS projects, so maybe we should just write: Sometimes)
you have to bind facts/a measure group, say cost correction data, that are physically unrelated to a particular complex dimension, say a product dimension.
[Sidenote: For complex dimensions, it is not uncommon, to have several, spreading hierarchies = relationship paths, like a tree rooted in the product variants.]
Now there are of course some logical and reasonable constraints which say that, e.g., the corrections are only relevant for products which are actually sold, listed in a particular part of the assortment,
etceterapepe ...
It is now quite straightforward to introduce calculated columns into the fact data source view which are defined by the approriate target attribute constants.
But when it comes to the measure group binding, however, there is just a single granularity attribute which you can choose (no matter what the extended attribute binding dialogue allows you to define ... the binding will only correctly work for the main granularity attribute).
What you really have to do in order to get this working is to introduce additional combinatorial attributes (whose keys are the combined key columns of the target attributes) into your dimension/attribute relationships (not into the target hierarchies). Then choose the combinatorial attribute as your binding granularity attribute, et voilà.