So, let's follow up Fredrick's post with some applied theory (and feel
free to flame/enhance my comments if needed, Fredrick, as my experience
as a winemaker isn't enough to be exact, but as a scientist/engineer it
shoulf be enough to knock off the rough edges).

Reality never reaches the full extent of theory. So if you take the
most efficient measure of conversion to alcohol, you know that that's
your maximum alcohol content possible. Use brix measurements, and the
further past 1.00 you go, the closer to the theory you get. What
doesn't convert, well, that's the yeasties being inefficient and lazy
(One can almost hear the yeast equivalent of Donald Trump swimming
through the fermenter, sticking out his pseudopod, and squeaking
"you're fired"). The "fudge factor" you're looking for if actually
hidden in the published differences between the conversion rates: the
spread between a 17 grams and 21 grams of sugar for 1 vol.% of alcohol
in a liter is likely the practical range the different authors (and
their sources) have found for their real-life experience.

What this really means is that if you "design" your winemaking process
to make a % alcohol at the most efficient rate, you'll likely end with
a wine with a slightly lower alcohol content than you thought. It'll
actually make it easier to make a "light" wine. I've personally given
up on actually measuring final alcohol content, as, being cheap, I
can't find a good simple measurer (I've never run into a vinometer that
told me that distilled H20 was actually 0% alcohol, for instance,
although that may be how they sterilize that stuff... Hmmm...), and
now I just compute expected alcohol content from starting sugar content
from my hydrometer, and I might round off 0.5% alcohol if it were to
stop at .990 instead of .980.

Hope that helps,and any clarification I need from those who know even
more about this than I do, please hop in.

Rob