Hmmmm…… do you have the original booleans where it is more smooth? One idea would to implement the shrink wrap approach on the troublesome areas and move the base mesh. Kent covered this for @adrian2301 during the 2nd or 3rd live stream. I think that this will probably could help solve that.
Or on a more manual but more controlled approach you could reference the booleans OR duplicate the red part and close it up and snap the green section to that smooth duplicate perhaps? Treat it like the Zerk fitting that Kent also made. He was using booleans for that workflow and then snapped the retopology to the simple boolean shapes (livestream 3: timestamp 2:57:00)
