Bias correction - normcdf and norminv giving 'Inf' values

I am correcting a simulated daily time-series of temperature using observations, then applying this correction to the rest of the time-series for which there are no observations. This is done on a monthly basis using distribution mapping.

However, when applying the correction to simulated values 'Inf' values are produced where the standard deviation is low (e.g. < 0.3), which causes the cdf value generated using normcdf to reach 1. I would like to find a way around this - I'm not sure if it's something wrong with the code itself or if I should use another function.

Here is a section of the code. Prior to this, the mean/sd of observed and simulated were calculated producing the attached matrix (mean_obs, stdev_obs, mean_sim_ref, stdev_sim_ref).

   %read parameters from calibration BC process,
           mean_obs = zz(mm,1);
           stdev_obs = zz(mm,2);
           mean_sim_ref = zz(mm,3);
           stdev_sim_ref = zz(mm,4);

            n = numel(Tas);
            tas_cor_ref = zeros(n,1); %pre-allocate x_cor_ref

                for a = 1:n
                    %calculate the cdf of each simulated value using 
                    %are the mean and standard deviation, obtained from zz

                     cdfval = normcdf(Tas(a),mean_sim_ref, stdev_sim_ref);

                     %calculate the inverse of this normal cdf
                     %x_cor_ref are corrected values
                     tas_cor_ref(a) = norminv(cdfval,mean_obs,stdev_obs);                                      
                end