Data Trees and Nulls / Wish

I originally tacked this on to the end of another thread but I think it got lost there at the bottom. 

It seems to me that the default behavior around trees and nulls is in general fairly counter-intuitive, especially with intersection components. I can't think of any situation where having the nulls in {A;B} and the successful intersections in {A;B;C} would be useful. I always end up having to resort to any number of weird tricks to get them to line back up. If I want the same definition to work with different inputs, where in some cases only valid intersections are produced, and in other cases certain intersections fail, I'd like to be able to count on the data trees being consistent, whether they contain nulls or not. It would make a lot more sense to me to have the results from an intersection always go to the same branch, whether or not the result is valid.

Here's a more concrete example. Say I have a definition with some curve intersections and I want to label the intersections of the curves. When all intersections occur, the labeling works fine, since the paths all match up. However when some intersections fail, everything goes haywire, and the data tree of the intersection results needs some major coercing + manipulation in order to get it to match the structure of the labels. 

Is there a reason why nulls get grouped in shorter branches, while successful results get separated into individual branches? My feeling is that the default behavior ought to be changed, but perhaps I'm missing a case where it would be beneficial. Also, I'd be very interested to hear clean workarounds for this type of case if anyone has them; I have a few of my own but they are pretty hack-ish and don't always work consistently.