I have read the study (DOI:10.1016/S0140-6736(10)61462-6) and, although it's vaguely written, it seems to be estimating total society-wide harm given current usage rates. Statements from Nutt such as "Overall, alcohol is the most harmful drug because it's so widely used" (reported at https://www.bbc.com/news/uk-11660210 regarding this study) seem to confirm this.

This is, in fact, the only sensible way to compare the relative harms of drugs. Otherwise, you can't put a number on how harmful X is without specifying the quantity and frequency of use, and they vary widely. With alcohol we know that lots of people consume small amounts weekly with little harm, and some people consume large amounts daily with major harm. The same is true for all drugs. The only data we have is for the current usage levels in society, so it makes sense to report harms for those levels.

At any rate, the paper boils down to "We asked a panel of experts how harmful are X, Y, Z in various ways on a 100 point scale and we averaged those numbers using some weights that seem vaguely plausible and here are the results!" And somehow the Lancet published it.