On sites that do not show mucked cards the EV graph is meaningless since when you go allin and are called you only get to see your opponent's cards if you lose at showdown. Therefore the -ev instances are counted and the +ev instances missed. (When someone else goes allin and you call you always get to see your opponent's cards so this is ok.) As a result on these sites, over any meaningful sample, it will appear that you have run significantly under expectation.
I think it would be better if on sites that don't show mucked cards the allin ev when you go allin and get called is set to the $ won in the hand, ie it is assumed that you run exactly according to expectation over these hands. The ev when you call an allin can be left unchanged.
If this is not possible to impliment then a simpler solution would be to just have the ev for all hands on these sites set to the $ won. Then the consolidated ev graph over all sites (some of which show mucked cards and some of which don't) would still be valid.