The most intuitive way would be to Tag a node before Out(child), so at the end of traversal last node’s parent will remain. But it will add a significant overhead in case of deep traversal, so it might be better to SaveR('child','parent') at the end of whole query.
Notice how 2 / B’s pid is 2 in the first query and 1 in the second?
From my perspective, the correct result is the second / when Save is removed. I don’t see how 2/B can have a pid (parent id) of 2 as per the data set or why it would change depending on whether or not name was saved.