Data collection on deaths began 5 years after collecting the BMI to try to limit this sort of effect, as they state in the summary. And they do a cause of death breakdown which eliminates the idea of a muscle-wasting disease.
60% of the excess mortality at low BMIs in the study was down to respiratory disease, most of which is due to smoking. Being skinny is a marker for being sick and for being old - both of which tend to come on before death, but neither of which are necessarily confined to the last 5 years of life. This kind of statistical association between being skinny and being ill is thought to explain most of the effect of increased mortility at low BMIs.
The question calorie restriction practitioners want the answer to is: "what is the effect of low BMI caused b...
An article about the pitfalls of using mice for animal research leads off with the example of calorie restriction. The controls for calorie restriction mice experiments were obese mice, suggesting that the health benefits of calorie restriction might be conflated with the health benefits of not being obese. I get the impression that people who study calorie restriction still think it worthwhile for life extension, but it's useful to try and integrate all evidence you come across.
[edit] timtyler suggests this is a well-understood effect that's already been taken into account by CR scientists.