I'd treat the graph of GiveWell's money moved as evidence in favour of meta (and in particular CEA) being promising, under three assumptions:

Yes, that and the stats for Giving What We Can/CEA look pretty good.

CEA is doing something orthogonal to what GW are doing (otherwise they might just be needlessly competing with each other)

I think competition tends to be good! It keeps people on their toes, and provides a check on problems. Consider your other point:

GW's top charities really are significantly more effective than what people would otherwise be giving to (otherwise that graph would just show the amount of money uselessly moved from one place to another)

With competitors you could check the rate of concordance, when they disagree, or look to see which organizations identify problems with data first, that sort of thing.