Assembly scoring ties, how to treat them?

In some cases (e.g. 5b1l) we can have assembly scoring ties. At the moment they are solved by assigning bio to the smallest assembly (see #166). Should it be the largest? 

Notice that in the 5b1l example there's another independent issue, whereby protein-DNA interfaces have 0 score (this is issue #161), which then results in 0 scores for all assemblies 