The study controls for income, so in theory the effect of lower socioeconomic status is removed. It is effectively saying fewer Hispanic people at a certain income level will die relative to people of other races at the same income level.

This is a paradox especially if your prior is that longevity and race should be unrelated.