I've done extensive tests on this on the TB20.
A lower rpm (2200) is about 10% better than 2400.
Peak EGT is 10% better than 100F ROP and about 20-30% better than flying full rich. Going 25F LOP is only 1% better than peak EGT - within measurement noise really.
Interesting, the engine performance curves for my IO-550 (different engine, but the same theory!) show significantly better economy at 25F LOP compared to peak.
One thought: in making those measurements at peak EGT and 25F LOP, IO540, were you using the richest cylinder or the leanest? If you have GAMIjectors (sorry, I can't remember if you have), the spread between the two may only be 0.5 USG/Hr. If you don't, spreads are commonly in the 1.2 to 1.8 USG/Hr, and the 'average cylinder' is actually running quite LOP when set up for peak EGT of the richest.