Pedantism, mostly. Indeed you have the right of it. Ethics can be murkier than one would like to believe. I have not yet seen a compelling statement on the matter of dichotomy with respect to ethics presented here, and semantics won't suffice.
Murky scenario: You run a clinical trial in patients with cancer testing the benefit of maintenance therapy with a drug for first-line therapy. Do you include a placebo arm?
If you do, then some proportion (usually randomized 2:1) will receive nothing new, just the current standard of care. This is bad for the placebo patients if the drug in question works, because they are not being offered the best possible care.
However, if you don't, then all patients receive the new intervention. 1% of the patients die of a bleeding event, say. Is this due to the drug? This isn't so hard to figure out, because clinicians are experienced experts who can make those judgments. But does it improve outcomes overall (ie, survival, quality of life)? You don't know. Ergo, ethically you are running experiments on people with a serious risk of not finding out the answer.
This is not a clear cut case of ethics, where one argument is 100% ethical or not. Both compromise ethics toward an ethical end. If the non-randomized study demonstrates clear, strong benefit that is unequivocal, then it is ethical in retrospect. If the trial drug kills people, then it was unethical. By definition, the latter therapy was not beneficent, and you didn't even design the trial in a way that could definitively show beneficence.
If this defies a textbook definition of ethics, I agree. But I also don't care, because there is no strictly ethical way to approach cancer research. The doctors and researchers cannot guarantee that what they're doing is in the patient's best interest. Therefore we must accept some murkiness.
I view the ethics of an uncontrolled trial as a mixed bag. Offering placebo would kill people, but having no way to know for sure means that a promising therapy won't get to people faster. And you can bet that the regs can recommend and have recommended that a randomized study be performed. This is what Vincent Devita criticizes, especially in the case where drugs and regimens have demonstrated unequivocal efficacy. A randomized trial in the case of a MOPP regimen in the 60s or in the case of obviously effective targeted therapy is entirely unethical.
AI
