The benefits of Scrum is not really Scrum itself, it's to have a standard by which to measure oneself. If you can say "this way is better than Scrum", great! But if you don't know any kind of methodology and are just winging it, you would be better off with Scrum 99% of the time, because it does, at the very least, enforce discipline through routine. Companies who don't have a measuring stick are likely to spend wayyy too much time on procedure tweaking. Speaking from experience.
Having this standard also allows you to tell stakeholders "actually, this is normal for this industry and I can prove it", which is invaluable when they're not very technically minded.
You're not really addressing the claim, which is whether or not Scrum actually works, i.e. is an improvement over any other method. I am not aware of any empirical, experimental or theoretical justification.
Anything you currently practice can be used as a yardstick. However, I don't think that practitioners should be spending time on this research. That IMHO leads to exact problems that the article whines about.
The idea that retrospectives are akin to scientific method is laughable. Maybe in 18th century that would fly, but there is lot more to scientific research than "continue doing what works and change what doesn't" (at the very least, following observations without theory leads to cargo-culting, which actually sums up the overall Scrum experiences pretty well).
I'm not addressing your claim because I suspect it's not the main reason people use Scrum, in practice. I'm saying it's a political tool, and any debates around its "effectiveness" are pointless as the real benefit is keeping management away. The efficiency gains that you discuss would be pennies on the dollar in comparison.
There is no evidence that Scrum is better than informal SDLC. If it was possible[1] to scientifically compare them, my guess is that Scrum will be net negative.
--
[1] unfortunately it's either impossible, or very expensive. I've seen cases at large companies when the same project was handled to two different teams as a competition, but in that case they used different technologies, not different SDLC methodologies.
Having this standard also allows you to tell stakeholders "actually, this is normal for this industry and I can prove it", which is invaluable when they're not very technically minded.