Not less likely to _have_. Less likely to be _diagnosed_. It is unsurprisingly difficult to be diagnosed with progressive cognitive decline in most places regardless of source even if you have been referred once for MCI.
I think the easier way of looking at it is this: there is a Lancet commission that produces robust characterizations of the modifiable risk factors that lead to dementia. If you're in a marriage that promotes reducing those risk factors, you'll likely have a lower risk of dementia. Isolating one variate like this raises interesting questions, but isn't a "smoking gun'-- especially in this case since the authors themselves indicate this is a surprising result given other studies.