Hoi,

pion->electron misid is not defined for arbitrary tracks (I don't know
of anybody who has actually looked at that number). It is only defined
for GTL.

If you  apply the track cuts  that are required  for electrons, you'll
see the misid rate drop to the permille level. At least I do. In 

  http://www.slac.stanford.edu/~ursl/gtl.eps
  http://www.slac.stanford.edu/~ursl/ct.eps

you  see the  true-pion-as-electron (left)  and the  true-pion (right)
momentum distributions  on the top, and  the misid rate  at the bottom
for GTL  selection and for CT  selection. While the misid  rate is not
yet at the  0.5% level, it is far from the  numbers Oliver and Henning
claim. I guess the continuing part of this story is to figure out what
happens with those tracks that are not GTL.

Applying  the PIDtables  on the  ntuple  level without  track cuts  is
entirely  consistent with  the  Beta-level Pidkilling,  i.e. when  you
apply track  cuts you will see the  low misid rate and  when you don't
you'll get the high misid.

We see marginal  changes (<1%) when applying PidKilling  on the ntuple
level with an implementation  completely different from the Beta level
PidKilling.

Just my two cents, your mileage will vary. 

Cheers,
--U.