To be clear: I don't think that anyone is surprised to see a 4% error rate here. No one ever really thought this system was perfect, and in point of fact perfect determination of innocence wasn't even a design goal of the jury system. The point to selecting juries from the public is to make it harder for a corrupt government to employ its own law enforcement apparatus corruptly. (Whether that works is also an argument of some topicality right now...)

But yeah: given that the error rate is in the 1-2 9's range, it seems like applying it to capital cases is a horrifying mistake. Imagine a medical device with only a 99% chance of not killing you.