Sure it's in RGB mode?? It's possible your display only "locks" to PAL and NTSC decoding is through software. If there were two crystals for the phase-locked loop, or both were decoded in software the quality would be pretty much the same. One color shouldn't bleed while the others don't unless there's a level or termination problem.Additionally, in NTSC mode, the red color is bleeding in a dither pattern, while in PAL it does not.
Got a 470uF cap? Maybe try that on composite/C-sync. It sounds like you do have a signal level problem. With a low level or bad time-constant/C-sync rise time the "back porch" area of the signal will be confused for sync which will mean longer sync = the raster lagging behind RGB = shifted to the left. Slowly over time the capacitor will charge up leading to a more stable image but it will never be right.So yeah, the problem goes away by itself after generally using the console.
If you isolated pin 6 and you still have color over composite (RGB not connected) there's something else wrong. And if there's still interference with pin 6 disconnected then I guess you have Bearking's problem.