So regardless of what alcohol does or doesn't do, your non drinking cohort includes all the people who are very ill and don't drink, which I assume is a larger effect than the people who are very ill and drink anyway (due to alcoholism or yolo).
Let's say 40% don't drink, 60% do.
If 1% of the population are terminally ill. Of those 90% don't drink, 10% do.
Then for 1000 people:
10 are terminally ill. 9 don't drink 1 does. 1000 is the total pop. Of those:
* 400 don't drink, and 9 of those are terminally ill.
* 600 drink and 1 of those is terminally ill.
Let's say terminally ill reduces life expectancy by 20 years.
The life lost per person due to terminal illness in the non-drinking category is 20 * 9/400, which is 0.45 years.
For drinkers it is 20 * 1/600 = 0.03 years.