Blank authors not ignored when calculating dominant author of work

In the RSI thread on this subject, Tim said "As mentioned somewhere else, it actually does this now—but that doesn't mean it will recalculate it without reason. Someone give me an example?" It's not always working, and I've listed several examples in that thread which I'll copy here. Could this perhaps be related to this issue, in which it appears that LibraryThing doesn't always select the most popular author for a work?

Examples from the other thread:

I combined http://www.librarything.com/work/3749902 (no author, many editions some with authors) and http://www.librarything.com/work/8229109 (author Gale Group, 1 edition). The resulting work http://www.librarything.com/work/3749902 still has no author listed.

I just combined works 1347611 (author Kathy G. Short) and 10602933 (no author), resulting in http://www.librarything.com/work/10602933 which has no author.

Another work where a perfectly valid author was ignored when combined with no author: http://www.librarything.com/work/3223970/editions ... works combined were 2163823 and 3223970

I've added a recalculation link (see http://www.librarything.com/topic/103802). I clicked it. Two of the three changed.

Basically, it chooses the most popular name, but it cuts the count of blank authors to a third of their original count.

The Landscape Quilt
http://www.librarything.com/work/3223970/editions Correct.
Not fixed. Blank has seven times as many copies as any alternative.

Research & Professional Resources
http://www.librarything.com/work/10602933
Fixed.

Worldmark Encyclopedia of the Nations
http://www.librarything.com/work/3749902/editions
Fixed.

This is great, thanks! I would have given a lower weight to blanks (10:1 in favor of non-blanks, perhaps) but this is fine.

I'll consider changes if cases are brought forth. Let's see how this goes, though.