Hmmm Not wanting to give away the reasoning utilized to determine the correct answer I am posting this comment - so as to receive automatic notifications of additions to the discussion of your QOD, which is some regards very similar to the QOD on SQL VARIANT that I had published on 10 Nov. 2011
Yes it is similar in some regards, but nothing like as similar as my Variant Order 1 question on 30 Nov. In fact I was quite worried when your question came out on 10 Nov, because of the very great similarity between that an the one I had submitted a few days before.
My series of three questions on variant order had its origins in the discussion of Paul's Collation and SQL_VARIANT QotD for 1 Nov, see this post
BTW, something has clearly gone strange with your 10 Nov question as the explanation has different column names, different values, and different number of rows from the question and answer. I don't think it can have been like that when it was first published, or I would have commented then (and even if somehow I hadn't noticed, I'm sure someone else would have commented on the discrepancies). Maybe you could give Steve a revised explanation (or maybe he wouldn't want one this long after the event)?