I think if you check from the beginning of 1999 to the end of 2013, the underperformance would be even worse than -3.6% per year. However, if you check from 1972 to end of 1998 and from beginning of 2014 until now, about 37 years, the largest market cap stocks performed much better. So yes, market cap weighted does not always perform better, but based on the past 50 years, it has been the way to go to get better returns, but also accompanied by higher concentration in particular stocks.

Hmm, that's not the way I would characterize the data.
Using the S&P 500 and predecessors since 1930, real total returns, rolling five year returns.
Basically the normal situation is that equal weight wins (69% of the time, and by a whole lot overall) but there are occasional stretches that the biggest caps go into fashion for a while.
Almost all of the five year intervals that cap weight had the advantage over equal weight were in just three stretches:
(a) The post war boom, generally five year stretches ending in the 1950s
(b) The tech bubble
(c) The stretches ending since 2017
Normally the few very largest firms are quite poor performers. But no rule of thumb is true all the time.

You considered only the data from (b) and (c) and the stretch between them, making the historically unusual look like the norm. Clearly this has been an exuberant bull market for tech firms, though at least most of them are much more profitable than during the tech bubble.

Jim