You want to bring a new drug to market, you should be required to demonstrate that it's not strictly inferior to existing options in at least some patients. I'm fine with a head-to-head that comes out a tie (competition is good for the marketplace) and I'm fine with a drug that only works in a subset if that subset can be identified. And I'm fine with a drug that doesn't work as well but is more tolerated. I'm not fine with a drug that loses in all respects in a head-to-head.
The FDA drug/therapy pipeline is supposed to give downstream users like doctors, public health officials, and patients more options within a certain risk profile. They're not supposed to be the be-all-end-all of treatment options.
If you’re allergic to drug A but not drug B, it doesn’t matter how much better A is than B. You need drug B.
All drugs have the potential to cause allergic reactions or other nasty side effects so unless a drug is too dangerous on its own, it should be allowed. It’s absolutely critical to deal with biochemical diversity in humans.
Alas, they are also more likely to fail (and give the competition data) so developers avoid them, at least for the initial approval.