A new computation of G from the Cavendish experiment

I used original data from the Cavendish experiment to compute the value of G. My computation yielded a value for G 2.67 times greater than the recommended value:

G(Cavendish) = 1.78424*10^-7 cm^3 sec^-2 g^-1

G(recommended) = 6.67428*10^-8 cm^3 sec^-2 g^-1

I used the well-known method of equating the torque of torsion wire (k theta) and the torque of gravity (GMmL/s^2) and solving for G:

G = k theta s^2 / MmL

k = torsion constant = 724.68 g cm^2 sec^-2
theta = excursion angle from mid-point = 0.0054788 radians
s = distance between weights = 21.97 cm
M = big weight = 158100 g
m = small weight = 729.8 g
L = gyration arm = 93.09 cm

More information about the geometry of the experiment can be found here. I would be grateful for any corrections and comments. Thank you.