Dave is correct, and the bottom of the drag curve does give the best approximation of the point for best L/D.
I found a very good method for calculating it exactly for a clean ATR 72 in cruise, here :
http://www.fzt.haw-hamburg.de/pers/S...n/TextNita.pdf
Page 32 for 3 different methods of calculating L/D and graphs on page 146. This confirms 17:1 as the correct practical value in clean configuration, with a theoretical value as high as 19:1
To get the best L/D in a glide it is neccessary to construct a vertical speed polar vs ISA, and then the best L/D is the ratio at the point where the tangent from the origin just touches the resultant curve. I would expect the 17:1 value to be still confirmed but the same AoA would occur at a much lower speed.