Absorbance is calculated as it's calculated. The detector records how many photons are hitting it when there's nothing in the beam and then it records how many photons are hitting it when there's something in the way. The difference determines the percentage of light transmitted. I'm not sure that fact has changed since the "old days".
A = log10(100/B) [when B = 100, A = 0]
The graph I created is a plot of B on the x axis and A on the y axis. This function is really only linear when B is in the 10-100% range.
If the detectors are still photon counters, I don't see how you can overcome this mathematical limitation without some type of artificial compensation in the software. Perhaps that's what instrument manufacturers do to increase the linear dynamic range? I don't know. My UV-VIS is about 25 years old. I generated this data a while back in conjunction with a project.
https://1drv.ms/b/s!AkH-uI0tnY5LddwtZJcO-6kGfAkThis detector is quite linear for the range of concentrations of the dye that gives a response no larger than A = 1.0 (middle of the visible region of the spectrum, pink line on second page of the pdf). You could probably extend it to A = 1.2 or 1.4 without suffering too much. It starts to go nonlinear after that. It's still predictable but it is decidedly nonlinear.