My normal answer would be no... Don't underexpose.
But I'm reading Mortensen these days and he of course solves the problem by adding light. In other words... Reduce the subject brightness range to 1:4 and you can then expose for the highlight and develop for the shadow.
According to Mortnensen, the four corners are better aim points because you will be successful at hitting them. So for interesting negatives, pick a corner:
Underexpose and underdevelop
Underexpose and overdevelop
Overexpose and underdevelop
Overexpose and overdevelop.
But according to Mortensen, don't try by any means, to properly expose and develop. That's boring.
If you read between the lines of Mortenson's ideas, it's basically N+1 method.
The important thing to remember I think is that when we photograph and print, irrespective of what's written before us or the theory that exists, we owe it to ourselves to do a few things to arrive at results we are happy with.
The by far most difficult quality to achieve, in my opinion, is tonality that suits and pleases the senses. I personally work SO hard to try to accomplish that. We need that before we can worry about anything else. Fine grain or not, if the print has shitty tonality, it won't be very nice to look at.
If we then go on a witch hunt, trying to minimize grain, we end up with things like theory regarding exposure and development. But to get contrast back up or down to what we like in our final product you introduce more grain at some point of the process.
1. If we underexpose our negatives and develop normal, we have to raise contrast when we print -> more grain.
2. If we overexpose our negatives and develop normal, we have normal contrast but dense negatives -> more grain.
3. If we shoot normally and overdevelop our negs, we lower contrast when we print, but because the highs are dense in the negs -> more grain.
4. Etc..
5. Etc...
We can't really escape it if we also want nice tonality.
The only sound method with respect to grain is to find the sweet spot between exposure and development, and then basically live with the grain the process gives us. It's such that if we try hard to eliminate one variable that we dislike, it's easy to introduce another that we dislike. Trial and error determines which consequences we can live with and which we can't.