Your response did not match my claim. Let me try again.
An operation - say, calculating n! - has a known complexity. This is not subjective.
It is a known fact that various sorting algorithms operate at different speeds. And if you are using bubble sort, it is very very much slower than the other common sorting algorithms. This, also, is not subjective.
I don't care if you enjoy super crunchy code - if you're only adding 2 small numbers together, and you've written a thousand lines of recursive bubble sort that takes days to run, it's bad code (even if it does somehow eventually come to the correct answer (and I don't think I want to know how you're using bubble sort to add two numbers)).
Now, that said, there is a subjective component; namely, *how* suboptimal something has to be before it is considered "bad".
And, here, there are many competing factors: speed, readability, simplicity, code reuse, maintainability, etc. Optimizing one is often done at the expense of others.
Contrary to the opinions of my more efficiency-minded (and, typically, bug-breeding) co-workers, most applications don't require the bleeding edge best - any standard form other than bubble sort is sufficient for most purposes. In fact, if you're simply sorting "the party" by initiative order / matching order / whatever, even bubble sort will work just fine.
So, yes, just as I talk about how balance is a range, not a point, and different groups will have ranges of different sizes, so, too, will different people have different ranges of *how* suboptimal things can be before they consider them bad, and that range will likely vary by the scenario. Which… is at least *related* to your "U-shaped curve", I guess, in that people will have different preferences. One can like chocolate cake or strawberry cake, sure. But if the chocolate cake carries warnings like major pharmaceutical products (may cause nausea, vomiting, diarrhea, blurred vision, seizures, or death), I'd say that's a bad cake, regardless of who likes it.
I like crunch. But I don't want *needless* crunch, I don't value "crunch for Crunch's sake". If you're using NASA supercomputers to track characters' every action, and probabilistically determine their inventory based on data the NSA has collected via spy satellites on what everyone was carrying at every point in time? Unless this is being done as a really expensive April fool's joke, I think you'll get better results by just tracking inventory than by asking the super computer whether or not you have a gun / change for a twenty / keys to the White House at the current moment.