It's not loss if you knowingly filter it out. It's call science and engineering.
As a science and engineering person I strongly believe that the human hearing cutoff is measured accurately enough that when applied to things like high bitrate mp3 compression the difference to loosless compression is minimum or near transparent, ergo most studies seems to support that hypothesis anyway, not many (or any) studies has been able to show otherwise that I know of.
The so call "frequency loss" is much MUCH less destructive to sound quality than, say, signal clipping from an overlty hot mix, which are much more easily heard. Instead of the industry pointing fingers at the format which are more than good enough in 99% of the cases, I'd rather they fix their mixing practices.