Thank you very much for your replies, Amb, Kvant, Majkel, Holland. I really appreciate it. I'm willing to learn. I've realised that the sonic impact of the ground channel is the same or almost the same as of L/R. Now I understand why my Sijosae AB amp with discrete rail splitter sounds so different from a similar amp with TLE.
I'm going to try the JISBOS buffer as active ground again, and add ferrites if needed. If there's still offset drift, I'll stick with it as rail splitter.
I'm going to try to use resistor deviders instead of TLE2426 to see if this improves SQ. I've seen your mosts on this before, Majkel, but I never uderstood how crucial this part can be. I'm going to try this in other amps like LISAIII clone and Mini3 as well to see if there's any difference. If it colors the sound sinking currents from the headphones, doesn't it color the sound sinking the small currents as well?