Just from a few snips of info, this dog sounds a little locked up in prey.
I think you have a few problems. Inconsistant correction methods mess things up. Stim one night, then no stim, prong but not hard enough. . .
The dog could have learned that whatever he fights through, eventually he will win. (i.e. the only time he outs is when the sleeve is slipped) When I used the EC for out work, that collar didn't come off the dog for 7 weeks in training. Every single time correction. . .every single time correction. Playing around with some correction here and there isn't going to help. If the collar isn't charged don't train.
Will the dog get after a decoy without a sleeve on? Just civil?
I think you need to go to a two pronged attack. . .or in this case a reward method and e-collar method.
Use a long line, very secure. Have the decoy agitate the dog at the end of the line. Then comes the bite and the decoy will fight with some intensity, letting the dog get lots of reward out of it. (what that means will be different for each dog)
The decoy will pull the dog out to the end of the line and the handler will secure the line and brace the dog as much as possible with his body. The idea is for the dog to be streched out at the maximum distance so he can't self reward on the sleeve. Countering, thrashing, pulling. . .
Then the decoy freezes, braces the sleeve with both hands if necessary and steps just to the limit of the line, without giving to much tension on the line. (you are not pulling the dog off, just at the end of manuverability) Keeping the dog higher on the sleeve helps, standing is more uncomfortable than not. Handler must move forward and control dog so it doesn't spring up and bite or anything else weird. Handler is leash man on the tie out.
The decoy stops all fighting and sits still, neutral, keeping the dog from self rewarding and basically boring the dog to tears on the sleeve for a good minute or two. (sounds like this dog will still be on the sleeve) You have to make sure he doesn't just drop off by himself without a command so you have to read his desire to keep on the bite.
Everything is dead calm and quiet. Handler then commands Auuuuuuuuuusssssss and on the SSSSS part after the long command gives a pretty serious stim. The second the dog comes off, the decoy breaks at a sideways run to trigger another bite (I always give the command to bite at the same time) while stepping into the range of the dog. Timing is very fast and critical, you have to reward the dog with instand animation and an rewarding fight and slip of the sleeve for outing. Then he carries the sleeve. At that point I would begin civil agitation and/or prey guarding to get that dog to forget about the sleeve and focus on the man. Have him drive the guy off and leave the field without the sleeve.
Eventually the dog learns two things as you progress. One, out means reward. . . two, out means avoidance of PAIN if done fast enough.
I progress with shorter outs and eventually outs off a still fighting decoy. . .still with the collar.
Later after the dog will out off the man with consistency you will have to move to a recall before he gets a second bite, or down, or what ever control methods you want to use to keep him from always making the immediate re-bite. Instead of just outing to get the bite, he has to out and down, or out and fuss, or out and ??? If he breaks those rules. . .stim before he bites and if he re-bites dirty no fight and stim.
The dog is learning two ways. The behavior that brings the reward, while at the same time the behavior that avoids the pain. Less conflict in the dog to out if he learns out means fight will begin again.
This is how I have done it, but using a prong. Now I do it like this using the collar with much better results.
Maybe someone with more experience can sharpen up this description for me, or give a better idea.