Only if you assume that (a) donors are actually aware of an 85%:15% split in the charities' disfavor; (b) approve of that. I would expect the naive assumption to be on the order of 90%:10% in charities' favor, but maybe that's just me.

Now, their donation pages for separate charities eg http://www.givewell.org/international/top-charities/villagereach/donate do state that the donation is direct to the charity, which is .a good thing.

So it's "I'm willing to take your money for me, but if you want to give it to X, give it to X directly" vs "I'm willing to take your money for me to split between me and X, or you can give it to X". Now on the face of it, that looks like X would get more money in the second scenario, as you point out. However there is an inherent naive assumption there that the split will be fair to X. If Donor A wanted to give $50 to Charity Navigator and $50 to charities through Charity Navigator, A has to give those amounts separately. If A wants to give $50 to Givewell and $50 to charities through Givewell, A may be tempted to just give $100 to Givewell under the assumption that Givewell will split it $50/$50. I suggest that donors who assumed that Givewell will be splitting at 50%/50% or better, have been if not deceived, at least permitted to operate under a false assumption where the one who could correct the assumption (ie, Givewell) benefits from not doing so. I think the split with potential breach of trust is more ethically dubious than the known split.

I'll admit that it's possible that Givewell have cleaned their act up since 2007. But they seem to have a significantly higher online profile than Charity Navigator, while also seeming to have a smaller number of charities rated and smaller amount of money donated due to their influence, which "smells funny" (or if you prefer, triggers heuristic estimates of suspiciousness) to me.