Now that I have an extra minute, let me fill in my comments a bit further...
The actual results you are quoting say this:
- Based on the limited data obtained, there is a 66% chance (1-P) that masks reduce infection of the wearer by 18% (1 - odds ratio) or more. Somewhere around 1/5 fewer cases.
- There is a 33% chance that masks reduce infection of the wearer by less than that.
- They wanted a higher than 66% confidence in their data (their threshold of statistical significance), and since they don't have that, they're saying they don't have sufficient data to make a definitive statement as to the effectiveness of masks beyond their stated conclusion that they believe it doesn't reduce transmission to the wearer by more than 50%.
- 50% was chosen here because the data would support protection up to 46% (the upper range of the 95% confidence interval).
They are saying the data is, more likely than not, indicating that masks protect the wearer, but there's a chance this particular data was confounded by sampling error, so their conclusion is "we don't know".
They then go on to say that their data assumes nobody but the wearer is using a mask, so this data says nothing about masks preventing transmission from an infected individual-- and preventing transmission from infected individuals is the reason medical professionals are suggesting wearing masks.
So this study tends to contradict your stated interpretation, but its authors don't feel the data is conclusive enough to make anything but a broad declaration and the test they're doing is irrelevant to the issue we are discussing.