The EFF report talks about how they got to 40,000 on page 12. That estimate makes several assumptions whose validity is tough to validate. I agree that it is probably inflated.
The 768 number is based on one release of documents, which only lists reported incidents. The FBI Inspector General estimated 6,400 violations between 2003 and 2006 based on a statistical sample.
Say the actual number is alot lower -- 10,000 for the period from 2001 to present. That would mean that like 98% of NSLs (about 40,000 are issued annually) were proper and consistent with the rules. But that 2% still means that there was probably 2,000 incidents where serious violations of constitutional rights were committed by people sworn to protect and defend the constitution.
I'm not clear on why your made-up number is better than their made-up number, but I agree that even a smaller number of that third category is too many.
Specifically, I'm hesitant to assume that the proportions of the EFF's sample are representative or that they 'scale up' to the number reported by the Inspector General and beyond. The main reason being while the EFF's sample is specific to violations recognized and reported as such, the Inspector General's report is specific to NSLs and not specific to things initially recognized as violations. The later report cites such comparatively minor violations as typos which are likely to be under-represented in the IOB documents, where more serious (and obvious) violations likely over-represented.
That is to say: there may be several thousand additional violations, but the proportion of them which are of the (arguably) more benign classes of clerical mistakes and "overproduction" may be much higher than in the EFF's sample. I think that 20% is very unlikely to be reflective of the (unknown) whole and I think the findings in the Inspector General's report[1] support that.
The 768 number is based on one release of documents, which only lists reported incidents. The FBI Inspector General estimated 6,400 violations between 2003 and 2006 based on a statistical sample.
Say the actual number is alot lower -- 10,000 for the period from 2001 to present. That would mean that like 98% of NSLs (about 40,000 are issued annually) were proper and consistent with the rules. But that 2% still means that there was probably 2,000 incidents where serious violations of constitutional rights were committed by people sworn to protect and defend the constitution.