In the digital world-
Some recording situations are more dynamic and require more headroom than others. You are in control and can either adjust headroom depending on what you are recording, or you can set things conservatively enough that you can use the same level targets all the time. The standards mentioned above are meant as appropriately conservative, agreed upon conventions.
Using less headroom is only potentially beneficial if doing so raises the noise floor of the room above that of your recording chain. You can’t do anything about the noise-floor of the room, it sets the lower dynamic range limit. The recorded signal needs to fit between the noise at the bottom and clipping at the top without ‘spilling over’ either end of the range.
There is no need to "leave headroom for post-editing". Once imported, editors use floating point calculations internally providing greatly increased headroom over that of the file. It is only whenever a file is stored that we need to make sure that the range of dynamics fits within the range of the file format.
To put it in actual recording terms, I'll relate a rather boring story-
I typically set levels on the Tascam DR-680 (writing 24bit files) so that the meters peak at the small line across the display, which is -18dbFS. I'm not entirely sure about the ballistics of the otherwise unmarked level meters on the DR-680, but doing that usually leaves a comfortable range of headroom. Setup for an onstage jazz trio recording, with some microphones arranged pretty close to the drums- about 2' away at closest, I had levels set so the meters were peaking somewhere over that line. I clipped the input a on a microphone closest to the drum kit, inaudibly, but indicated on the display. Many other times recording the same band in the same room, same situation, I made sure things were peaking right on the line and it never clipped. At the time I had been recording a bunch of other less dynamic stuff which didn't require as much headroom and had become somewhat complacent, letting the levels run a bit hot. It wasn’t major problem but pushing the recording level hotter didn’t benefit me anyway. Now I always target that -18dbFS line and avoid clipping.
I consider that situation (close to a jazz drumkit with an energetic, dynamic drummer) one where I'll use the most headroom. That transient rich source at close proximity has a high crest-factor or ratio of peak to average level. Recording a PA reinforced band from farther back in the room won't require that much headroom. The crest-factor is much lower and the transient peaks are less wild. I can safely get away with using less headroom in those situations, but I still use the line on the meter, because doing so simplifies setup, provides consistency and keeps me from wasting time guessing how much headroom I really need, and because I know from experience that the noise-floor of my recording chain will be buried far enough below the noise-floor of the room that I won't benefit from raising the recording levels at that point anyway, the noise floor of the room would just be raised right along with the peak levels.