Looking back at my 1997 GAMI tests on the TSIO-520 engines, I recall a number of factors that should be considered when finding roughness in the lean. During the test (you should have a pre-gami test as well), I measured airspeed, CHT and allowed a few minutes for CHT (and turbos) to stabilize at each fuel increment.  My tests showed 93% airspeed on 79% fuel at LOP for my aircraft (twin cessna) versus the ROP test.  During the transition to GAMI mode I ran into all the sins that are concealed by ROP engines.  I had to:
1) repair an open sniffle valve on the intake allowing air into the induction system in flight.
2) pressurize the induction to find leaks in the intake couplings and there were many.
3) plugs, wires, mags all had various issues, highlighted only when LOP measurements were made.

By looking at CHT you will find cylinders that may be losing power faster than the other cylinders.  This can help focus the search beyond just injector issues.   With a lot of detective work all these issues disappeared and the roughness was minimized to the point it was not an issue once LOP was stabilized. I recall GAMI also changed out one injector once we sent in the data and graphs.   I used that setup to fly the Atlantic without requiring aux fuel tanks.   I think I had somewhere around 7.5 hours endurance at FL250 in the Cessna on stock tanks.

My main point is to consider all those factors because most small leaks and problems may not show up as roughness in ROP operations unless the problem is severe.

good luck, I think it worth the trouble if you want better endurance or fuel economy.

Paul Miller
Legacy RG