There are two aspects of mortality / demography research that I think matter a great deal here: the factors that are controlled for and the aggregation level.

Reading the primary source paper reveals that they account for age, world region (so not only high income countries), gender, study length and a few other factors. They use random effects weighing to account for variability between studies. I don't know if they've accounted for income levels, but that doesn't undermine the statistical significance of the correlation between social isolation and mortality rates.

Secondly, most of the research is done in an aggregated way to describe general patterns. Of course there will be (systematic) exceptions to the rule, like zen monks; this research does not imply that social isolation will necessarily lead to a shortened life, only that a large group of socially isolated people will on average have a shorter life span. Demography research is I think about general trends and patterns and not so much about individuals or small-scale groups.