For example, one European and one Asian could make 10 European/Asian (50/50) mixed children. Some would look purely Asian or purely European, and some of a mixed phenotype in between both.
Eye shape differences are mainly a few genes. Plus let's say 3 kids have blue eyes (2 main genes OCA2 and HERC2) whereas the rest have brown.
Thus, 3 kids born with European eye shape and color, perhaps from random genetic luck or unknown environmental reactions(maybe the mother has lower vitamin d levels that then favours genes for less melanin as biological adaptations) could pass as European looking despite being half Asian genetically, while his brothers pass as Asian since they inherited those few eye shape/color genes from the Asian side instead of European, though are half European.
So, despite being 99.99% genetically similar siblings, that 0.001% difference in a few key genes in the eye area makes them look very different, the difference between Russian or Korean, perhaps.
Phenotype-genotype correlations can be an indicator at times, but can also be extremely misleading and downright incorrect at other times. It depends.