I have some concerns with the assumptions in generating these numbers.
First, I think it would be fairest to focus on only core residencies. The numbers are presented for both core and subs. If someone "washes out" of a sub it's not the end of the universe. They are still a board certfied physician who can practice -- perhaps not in their field of choice, but that's life.
Second, I have a problem including transfers in this total. First, as mentioned above, many transfers are probably by design. Second, even if not, these people got another spot. This happens in all fields -- someone starts in one company, it doesn't work out, and they move somewhere else. Regardless, their career is still on track.
Third, let's exclude those that died. This has nothing to do with being "washed out".
Last, let's assume that all of the withdrawns are really fired. This is certainly an overestimation, as some will be by choice.
This gives us 258+863 = 1121 residents who were "let go" by core programs. Compare this to the denominator of 25346 = 4.4%. (4.4% over 3-4 years, 1.5% per year)
Now, we can argue whether 4.4% is too high, too low, or just right, but I think it's a much more accurate number than 12%.
And, if we assume that of the withdrawn, 50% did so voluntarily (I just made that number up), the actual failure rate is 2.7%.
And, this includes all comers -- US grads, US IMG's, FMG's, DO's etc. It's possible that the "failure" rate is very different between these groups.