2017 trial about both viral and bacterial respiratory illnesses found n95 to be superior to surgical masks when looking at 3591 nurses in Beijing.
Surgical masks showed fewer infections than no mask, but it failed to reach statistical significance. If you click the full study (it's free on Wiley. I just linked the synopsis because I thought the "quick and dirty" would be generally preferred), you can scroll to Table 4 which looks at essentially all droplet-spread respiratory illnesses.
The findings?
Continuous N95 arm RR 0.26 (0.16-0.42) P<.001
Targeted N95 arm RR 0.43 (0.25-0.72) P=.001
Medical mask arm RR 0.65 (0.41-1.04) P=.074

What it means:
RR is relative risk vs control. The parantheseis gives the range that could be expected by the data above and the P is the decimal (convert to percentage by multiplying by 100) chance that this was all due to chance.
So for surgical masks, the study found nurses in sick patients were 65% as likely as someone not masking to get sick, test positive, or be able to culture the same pathogen the patient had. Analyzing that 65% closer says that you may be as good as 41% as likely vs nonmask but as bad as 104% (or slightly MORE likely). The chance that this is random chance (you roll a pair of dice looking for snake eyes and get 12 in a row. A statistical anomoly that very easily could happen but could never continue when carried out an infinite number of times...) is 7.4%.
Generally speaking, when a the number in parenthesis (confidence interval) crosses 1, it means that the available data are not statically significant to draw conclusions from. Similarly, the p value being >5% is generally higher than one would like. Doesn't mean you should just ignores the data, but it means that a better designed or larger study should be done.

So what does that mean for our n95 in the study?
Cuts the rate of transmission to about 1/4 nothing, and we are pretty sure that's not by chance (>99.9% sure)

Problem:
They're uncomfortable.
They can dig into your nose with extended use
If you aren't used to wearing them, breathing is fairly difficult. It's not a fat/skinny or in-shape vs unathletic type of thing. It's just an unusual situation and people take them off a lot.
Good news is that targeted n95 mask use was shown to be effective, but not as effective

The last point is that n95 masks must be fit tested or a general person's use of one isn't nearly as directly applicable.

And then lastly lastly, because of this and the propensity for it to spread via healthcare workers to patients, getting them into hospitals took priority to getting them to normal people and will continue to do so until there is a good supply so that the car wrecks who lay in a trauma ICU bed for a month don't get (and several other groups of similar long-stay patients) don't get COVID as well.