Re: 6 units/day = Max. Longevity. 13 == teetotaller longevity.
I'd love to see some references: not that I disbelieve you, but I'd like to be able to point other people at that research.
This study uses American 'standard drinks', which are roughly 77% more than the UK unit of 10 ml ethanol. The 'J'-curve in the article bottoms out at roughly half a US 'standard drink' per day, or one UK unit per day, which is quite a difference compared to the figure you state: however, one of the things that is difficult to take account of is significant under-reporting of consumption. While a reporting rate of 16% of actual consumption would be surprising, I would expect the effect under-reporting would be to shift the bottom of the 'J'-curve to the right.
As for under-reporting: the problem is well known - for example, if you total up how much alcohol is sold in an area, and survey the inhabitants for their drinking habits, you can get discrepancies of around 50% of the total sold. Efforts are being made to be able to allow for under-reporting, or mitigate it by using better survey techniques. E.g, a smartphone app to allow a subject to report consumption at the time of consumption; and population measures looking at blood biomarkers that vary according to alcohol consumption - there's an instructive paper comparing Finns and Russians.
When Self-Reporting, People Underestimate Alcohol Consumption by Nearly 50 Percent
The Validity of Self‐Reported Alcohol Consumption and Alcohol Problems: A Literature Review
Recent reports of consumption are validated more easily than drinking patterns measured in drinking practices surveys (as evidenced by coverage rates of surveys to sales statistics ranging from 40‐60%).
Participants reported consuming more alcoholic drinks during real-time assessment than retrospectively. For daily accounts a higher number of drinks consumed in real-time was related to a higher discrepancy between real-time and retrospective accounts. This effect was found across all drink types but was not shaped by social and environmental contexts.
Summary here: http://www.academicwino.com/2015/06/self-reporting-alcohol-consumption-app.html/
In RUS, elevated CDT values were observed in 36.6% of the men and 17.6% of the women. In FIN, the respective rates were 9.6% and 9.4%, which are similar to average European rates. The prevalence of elevated CDT values seen in RUS is the highest prevalence ever reported in general population surveys. However, the self-reported alcohol consumption was similar in the two regions. These results suggest that alcohol consumption especially in Russia may not be reliably estimated by self-reporting...