first, dont use a 1 to 1.
second, the training could be retrieved over assocation from the student object because it is probably in context, any deeper/further xpathing is not possible and should be handled by the attribute or serverside (microflow), usually if its not a native app i go to a microflow anyways for database retrieves, since youre going serverside anyway.
I think in a nanoflow, you will need to retrieve those by association, not from database.