At the bottom of the description of the kptbounds input variable - http://www.abinit.org/documentation/hel ... #kptbounds where the lengths of the segments is given explicitly for rhombodedral lattices, the relative length for Gamma-T is half what it should be following the convention that was used here.
There it's stated that l(Gamma-T)=1/(2*c) and l(L-Gamma)=sqrt(4/(a^2)+1/(c^2))/3
If we take alpha=60 degrees (a=sqrt(1/3) and c=sqrt(2/3)), these two lengths should be equal.
Instead this gives l(Gamma-T)=sqrt(3/2)/2 and l(L-Gamma)=sqrt(27/2)/3=sqrt(3/2).
So probably it should be corrected to read "l(Gamma-T)=1/c" so that the relative length in comparison to the other segments is correct.