Yeah, the thing about market cap weighted funds is that despite the company count in the fund the returns are mostly driven by a relatively small number of large companies. For example, even though VTI holds shares in more than 3500 US companies, the top 50 companies represent about 40% of the fund. So they're really a lot less diverse than you think, and they arguably have a strong winners bias where company growth doesn't really influence fund returns until that company has already gotten truly massive relative to all the rest.
That said, I do like how market cap weighted funds tend to be very low cost. A common problem with alternative index research (value, momentum, etc) is that the returns using a certain filter may look terrific in theory but the internal trading costs for buying and selling companies that meet the filter can exceed the ER by an order of magnitude. I did notice that the research in the OP does mention trading costs (which I find refreshing), although I personally think they dismiss the effects too quickly with not enough supporting data of just how high trading costs really are. Toss in tax implications of regular fund trading (especially in momentum strategies) and the higher average ERs of "smart beta" funds and it's even worse. So the academic returns often have a practical limit that the researchers do not understand but that traders would immediately recognize as a dealbreaker.
Personally, I'm fond of a relatively simple method of approximating an equal weight fund using two low-cost market weighted funds -- VTI and VB. By purchasing 50% of each, the total stock portfolio is pretty evenly distributed among large, mid, and small caps and also among value, blend and growth companies. Look at the Morningstar style box distributions for each, and you'll see what I mean. (
VTI,
VB)