My 2003 article on the race-of-victim bias claim is here.  A common theme throughout these studies is that an apparent racial disparity may appear in aggregrate, unadjusted data, but it disappears to a minor or even statistically insignificant factor when valid variables are controlled.

And, yes, county of prosecution is a valid variable. In Maryland, for example, the people of downtown Baltimore City elected prosecutors who sought the death penalty more sparingly than the prosecutor elected by the people of the suburban ring Baltimore County.  That "geographic disparity" is local democracy working as designed. That was their choice to make, although the result was fewer death penalties in black-victim cases.

The other interesting thing about the argument is that the opponents wax indignant that we supposedly value the lives of black victims less. In making this argument, they implicitly concede that executing the killer constitutes valuing the life of the victim. Game. Set. Match.