Part of the problem is what is "normal" is subjective and may depend on the
situation/context. "Small" stars are still "big" compared to many things,
for example. Additionally, statistically, the median is often more
reflective of "typical" amounts than the mean; many real-world values
follow a log-normal distribution or Pareto distribution, such that the
arithmetic mean is significantly higher than the median, and doesn't
reflect what a "typical" example looks like.

All these makes me wonder whether Loglan had it right embedding
comparatives (x1 is taller than x2) into its primitives / gismu :p