Skill Design - opportunity cost

I was remembering american pie and various other comedies where characters would flaunt their ability to observe undetected, their target gender. I was posing the question here and later here for more serious tactical discussion.

It brought me to some interesting mechanics design problems

Perception is more bang for buck at 5pts per level vs 4pts on Observation. And there are loads of other skills that give context analysis or allows the character to get in close.
Why by a skill when defaults have very little penalties but it takes 200 hours per cp of skill? (like running and jumping, and there are other skills that have that conflict)
If Observation is getting close to the target and tactical context why have so many cheaper and better solutions:

Getting close - shadowing, acting, stealth, camouflage, streetwise, savoir-faire, pan-handling
Context Analysis- Guns+Per, soldier, "warrior profession", strategy, tactics, Information Analysis, cold read/aka fortune telling, body language, psychology, expertise relevant to the situation

I realize that if there is a skill being designed, we need to look further at the opportunity cost of mastery. it the skill has no point over mastery - like observation then maybe we should consider it as a sub-specialization or another feat within another skill.