To a certain extent they can be, however generally speaking the very tops of this set tend to have both good luck, good genes and good lifestyle. It's hard to get to right near the very top of lifespan without all three, and the population from which we select centrenarians is extremely large. Statistical blips (good luck or unmeasured factors) can be relevant, even highly relevant, but unless you're overlooking very important things it's extremely unlikely to rise to the top without favorable outcomes in other factors.

So maybe we observe things that are 95% of optimal instead of the best genes in the data set. Or maybe we observe diets that have 90% of the performance of optimal diets. We still learn a hell of a lot.

By optimal I mean best-performing in terms of life expectancy from among the total data set (people) not just among centrenarians.