There are several issues that concern me here:
Operators do not strictly follow the 3-pan audit process well and parts rejection are caught in the final audit stage...
I am not sure exactly what you mean by this but if the 'scrap' parts are being caught in a final inspection this really concerns me. Inspection should be a double check not an actual check. One of the things I have heard and agree with is "100% inspection will not catch every part".
Mostly I would ask how the inspection process is being performed? Is it mechanical (say a camera system) or human interaction (human with a measurement device). Also, if parts are found in final inspection, would this not trigger some type of action "why did final inspection find bad parts".
the TPY is actually about 85%
The output is running at 85% and 5% scrap. Personally, I would worry about reducing scrap then worry about the 85%. I don't know the process you are running or the qty of parts run, but 5% scrap (to me) seems a bit high and should be the first 'fix' (if possible). I say if possible as some older machines just cannot hold some requirements. Are the machines capable of running the parts to the requirements?
Are the requirements acceptable? For example, long ago I had heard of something call 'SPC creep', where each year an improvement of CPK was required. After 10 years or so, a tolerance of .005, an easy tolerance to hold, became .0003 because they had to improve. Are the tolerances reasonable to your process?
The setup time takes more than it was set up in the system.
I have seen this as a common issue, we experience it here as well. We are addressing it in two ways. 1. We are looking at the setup time and asking 'is this realistic?' and 2. is there something we can do to speed up the setup time.
For question 1, if we give 1 hour for setup and everyone takes 3 hours (or the average is 3 hours) we change the setup time to 3 hours. On question 2, we look at cost vs. benifit, we setup a part once a year and the cost to speed up the setup is $500. We don't perform and suck up the 'extra' time. If however, for $500 we can save 25 hours a year then we will do so.
Just my thoughts, hope they help.