Large parasitic losses are indeed an engineering challenge. In theory, you are correct. However, you are making the assumption that efficiency is a static num

Large parasitic losses are indeed an engineering challenge. In theory, you are correct. However, you are making the assumption that efficiency is a static number. In reality, parts age, get dusty, out-of-alignment, etc. Let's say you are building something that makes 10 units of power, but 8 of those units are needed to keep it running (when working as designed). Say that the part that uses the 8 units of power is nominally 90% efficient, but after you've run it for a few months it drops to 80% efficiency. Now you need 9 units of power to keep running, cutting your output in half. This is not just a problem in terms of dropping output; it is also an issue from the standpoint of understanding your machine. It's hard to engineer something when small changes have large effects. Your output just dropped 30% - is something broken, or is the fan just dusty again? How long do those electrodes really last anyways? Did someone use the wrong cleaning fluid? Maybe the pressure gauge is just off by 1% again. All those things add up, just like coding a big ball of mud can make debugging difficult.