We revisit the measurement of the polarization fraction, P , and the measurement of the polarization angle of partially linearly-polarized gamma rays using their conversion to e + e − pairs in the field of a nucleus. We show that an inappropriate definition of the azimuthal angle, ϕ, used to reference the orientation of the final state degrades the precision of the measurement of P , by comparison to the optimal case where the bisector angle of the electron and of the positron momenta is used. We then focus on the lowest part of the energy spectrum, below ≈ 10 MeV, where a large part of the statistics lie for a cosmic source. We obtain the value of the polarization asymmetry, A, of pair conversion at threshold and we show that in the case where the correct expression is used for ϕ, the measured value of A tends to the limit.