Why only transcriptomes vs protein databases?

[jefdaj]

I'm trying to find orthologs in uncultured species, but it only works with transcripts and protein databases. Is there a reason it couldn't do other combinations? For example in my case it would seem natural to start from proteins as the query and a genome as the database. I'd be willing to make some changes but don't want to if it's either biologically wrong, hard, or not worth the performance gain vs just using crb-blast. Both seem at first glance like they wouldn't be too hard: for protein query skip the 6-frame translation step, and the command for generating the database would probably just need an argument changed somewhere to use DNA?