Another problem is that, when we hypothesize higher level mutations (which pretty much have to be hypothesized), you run into two problems: selection and serendipity:
Selection: Higher-level (regulatory type) mutations are more difficult to provide selectable fitness improvements. This is because they are advantageous in certain circumstances. I’m not saying this is always the case, but often, selection is going to be weaker. Of course you can call upon drift, but drift is a much weaker search engine.
Serendipity: If you get a human from a primitive, rudimentary, small, primate via higher-level mutations, that means all the underlying, foundational, hardware was already, a priori, in place, for the higher-level mutations to take operate on, and take advantage of. Its like having all the parts of a car constructed and assembled, and then a few random mutations supplies just the right screws and bolts, at the right places, and you then have a car. There is a lot of serendipity implied.