This is a bit of a controversial topic among many people I know--not settled by any means--but portfolio theory doesn't really say what you think it does here. In particular, portfolio theory suggests diversification because it reduces variance, generally at the cost of some (small) amount of return. _Since your utility from money is concave_, variance reduction improves expected utility.
Your utility from lives saved in the third world is (or should be), I think, linear. (Perhaps even convex, though that's a much more complicated argument.) While if I were a doctor, the 100th life I saved would seem a lot less interesting than the first, I don't think it has any less moral impact. Conclusion: variance reduction has little to no value, and if you think charity X is (in expected value) any measurable amount more efficient at $goal than Y, you should reallocate all donations from Y to X up to the limit of what X can efficiently spend.