For example, someone in a truly blind rage might go ahead and recklessly kill someone with complete and total disregard for his own life. While observing this act we might say that he does not believe his life to be precious because he acts as if it's not, yet he can nevertheless still be fighting and killing someone because he's driven by dark emotions.
However, the logic should be consistent for any bounded cutoff.
You could construct a system where everyone with lower than given value is subjected to some action. Given your (or societal) bounded knowledge you then cannot generalise because you are not omniscient. This means you get to evaluate everyone on their terms. This view is incompatible with stereotypes.
Alternatively you just assign values by axiom, which will likely end up with a big and likely inconsistent philosophy, unreliable guide.
Indeed egoism like this is a valid strategy but not an optimal one in absolute terms. (Tit for tat is stable instead over time.) A philosophy or strategy should be viable in the presence of a mixture of them.
Survival instinct is powerful, it doesn't necessarily follow that exercising it means that you consider your own life precious.
Otherwise you presume either hard equality between everyone (easy to rebuke) and/or omniscience.
Or you revert to consequentialist ethic where only consequences matter. (Not essentialist.) From that place, going from basic utilitarian ethics you can show what sort of value certain human lives had.