Because to some extent the asset classes are uncorrelated. This is the basis of Modern Portfolio Theory, and the reason you don't just dump everything into the highest-return class (e.g., stocks).
Maybe. More asset classes shouldn't hurt, although at some point they may not help much (new ones may be highly correlated to some combination of the previous ones) and lower fees are nice. For bigger portfolios WF's single-stock approach may add enough value to compensate for these. I'm sure both companies have simulations proving they're better...
For a simplified model, check out Kelly gambling
https://en.wikipedia.org/wiki/Kelly_criterion