I think ultimately it comes down to the targetting (min/max diff settings), the diff1 mask, and the multiplier required.

Scrypt-N pool settings were using the Scrypt Diff1 Mask. Pool target of 16, Min of 16 and Max of 1024.

Lyra2RE settings we are now using the Quark Diff1 Mask (Scrypt Diff1 mask seems to put LTCMAXMYR's rate at around 6gh/s which is obviously wrong also) and the X11 Diff1 Mask results in too many "share above target" settings. Min of 0.1 and Max of 6.4 (based on the 16 to 1024 ratio).

If Hashlink and others can post their settings, we can maybe iron this out.

I don't know if the difference is because of ccminer and sgminer however, but there is something weird.