So I think I have found the issue with the funky graphics. Both the 74HC244 and the GAL16V8 turn on at the same time and try to send data to the RGB output at the same time. I think I have something wrong in the GAL programming. The /OE is not doing what an /OE is supposed to do. Which is probably a mix of me not knowing much about GAL programming and a limitation of the number of registers in the 16V8. Have the slightly bigger 16v10 coming so going to try that. And thanks to a fellow Hackster I have some 6331-1 PALs and a programmer coming. This is what was in the original Excalibur 64 used. I avoided using it my recreation because the 6331-1s are expensive, can only be programmed once and need a special programmer. So yeah now that I have the opportunity to do both I can see what works. But if this is going to be a going concern I think I will be leaning towards the GAL16v10. Simply because is available and can be reprogrammed, can do stuff like play around with different palettes. 

There maybe issues further back in the circuit either noise or issues with mixing different 74xx technology's, 74LS, 74HCT etc... as some of the waveforms are not very clean.