Exonerate still is pretty much the same. But, you're using your whole genome as a query. This is useful only when you have a LOT of thing to search and wants to reduce memory IO. Exonerate always return the best match unless you explicitly specify a "suboptimal threshold" (e. g. using --percent).

Anyway, there's a possibility that your case to be an output artifact related to the geneseed threshold, fixed in the 2.0.0 -> 2.1.0 transition. Just check the changelog at Exonerate's source.

Google returned me the man page here. You can/should specify -q and -t. But why use this tool? When fastas grow big I am almost sure blast+ to tabular output or limiting its output to the number of hits you expect will do a faster multi-thraded job directly on fastas as well.

My 2ct