Say I have the entities 'sale history' and 'sale forecast'. For a given product and period, I want to see the actual sale versus the forecast sale. However any given period may not have had a forecast or an actual sale, so I use an SQL like :
SELECT NVL(f.product_id, h.product_id), NVL(f.period, h.period),
f.forecast_sale, h.actual_sale
FROM forecast f
FULL OUTER JOIN history h ON h.product_id = f.product_id and h.period = f.period
Basically I am joining two child tables together on a common key. Full outer joins are rare in relational databases as normalisation would generally merge the entities with common keys. Left and right outer joins are more common as the typical use case is to select the parent and its children while requiring a row even if a parent has no children.
So if you have a full outer join, the first thing to examine is whether the data structure is correct.
The data structure in an object model is fixed or 'pre-joined'. If the data structures within the model don't readily support the production of a certain result set, then you pretty much have to go without (or at least code up a lot of functionality to extract and join the data manually).
If you post some details about the relevant data structures the advice may be more precise.