In line-of-sight massive MIMO, there is a nonnegligible probability that two users become highly correlated, which leads to a reduction in the achievable sum-rates of linear precoders. In this letter, threshold values of a previously proposed dropping algorithm are found analytically to avoid repeating a large number of simulations to find the optimal threshold. These thresholds allow us to improve conjugate beamforming (CB) and zero-forcing (ZF) sum-rates with max-min power control. By using the proposed threshold values, the CB and ZF sum-rates are maximized, when there are only two correlated users. In addition, by using the derived thresholds, a modified dropping algorithm is proposed for channels with any number of correlated users. The results of the simulation scenarios show that at signal to noise ratio of 20 dB and 120 antennas at the base station, the modified algorithm improves the average CB and ZF sum-rates up to 36% and 5%, respectively.