@silentguy - it sounds a reasonable request, though I suspect it would be a major problem to implement with the way the comparisons work. From memory it computes a hash using a combination of title and author (depending on the options you have specified). If you do an "ignore author" search, then the hash it generates is done only on the title. However what you are asking for would not fit into this scheme, as you can't mix title only hashes with title/author hashes.
I will take a look at the code sometime to confirm this but I suspect it will be a non-starter because of this. I would have to code up a huge amount of special case logic and given in two years this is the first time this has come up I would rather not do that.
I can only suggest you use the "Ignore Author" searches if you are trying to find duplicates based on title only. Having the author field set to Unknown is not a particularly common thing for calibre users to do, certainly not for the bulk of their library
