>>7632270
As you figured, the head models don't map to a strictly 3D model. That's why it's not going to make sense when you try to apply rigorous 3D construction logic to it. Basically the head and face is drawn in the most appealing way (to the artist) for direction it's viewed at. For example, you'll start to see things like people drawing the ears closer to the face on a 3/4ths view, which helps keep the face from looking too wide, particularly in the absence of more 3D looking lighting. The way you learn the "nuance" is by seeing how other people do it and figure out their "rules" for every view. There are not many of them, and you could mostly split up the problem into front, 3/4s, side, back, and combine it with above, below, and leveled. Any in-between angle could be realized as a mix of the prior set (interpolating). I think what would help you with models is looking at settei, which animators use to stay on-model and determine how the character looks from the common angles.
Some strike witches settei: https://www.moehui.com/206718.html
Repository of other settei: https://setteidreams.net/settei
Pic: Cardcaptor Sakura settei