That can be adjusted out. The program does notch the gain up at high frequencies, presumably to help you catch some of the consonants that might otherwise get lost in noise and perhaps to deal with the change in frequency characteristics of speech that occurs when people raise their voices. But one could drop that high frequency gain back down and try to make comparisons between identical gain settings with wide versus narrow directionality. It’s hard to find these numbers. I can’t, off-hand, find any work that has compared narrow to wide directionality specifically. It seems like they compare narrow to omni, or wide to omni, in different studies which makes direct comparisons more difficult. But it looks like in the lab, wide directionality will give a 2-3 dB SNR advantage (closed fit, only ~1.3 dB advantage for open fit) over omni and narrow might give a 4-5 dB SNR advantage. In the lab is important, because it is always an artificial situation. There has been at least one small study trying to bring a measurement task into a natural setting that has shown that, with an open fit, there is no significant directional benefit. It has not been repeated yet with a closed fit. Plus, while it might be logical that the narrow directionality benefit will be additive in all cases, it could just as well be the case that it really only benefits a closed fit. I don’t think that comparison has been done. Perhaps, though, we could say that narrow directionality (one of the big differences between the 30/50 level and the 70/90 level) might give 1 dB advantage in noise in a natural setting. That seems like a small number, but in the right environment it can be critical. However, there are certainly situations where the signal to noise ratio is so poor, or the individual’s SNR requirements are so high, that this won’t make any difference. So it might provide a critical benefit, but only in the right place for the right person. Further, it has been found that people are pretty poor at managing their manual programs, so you might have a higher chance of taking advantage of that benefit if it is automatic (i.e. 90 level).
How can an individual quantify that benefit? A small boost that might be critical in a particular situation that may be individually unpredictable. It almost has to come down to a gut feel–your back brain data mining your own experiences. (And as soon as your button-pressing finger smooshes some dirt into that back mic, all directional benefit is lost anyway.)
As an aside, one of the limitations of the sort of research that has been done is that to create these artificial speech-in-noise situations they simply raise the volume of the background noise but the recorded voices stay the same. As I mentioned, in real life situations where people are raising their voices the shape of the speech changes. As you can see, this research is complex. The R&D is complex, for anyone who thinks that modern hearing aids are simply 60 cent amplifiers with a billion percent mark-up.