Sorry, it's the "except after 'c'" bit (which I left out) that's mostly incorrect:

  # cat /usr/share/dict/british-english-insane | grep -c cie
  1375
  # cat /usr/share/dict/british-english-insane | grep -c cei
  352

So that works only ~20% of the time.

With the general "i before e"... Is being more often true than false enough to make it a valid "rule"? Using your numbers it's correct ~60% of the time, with my dictionary it's correct ~76% of the time:

  # cat /usr/share/dict/british-english-insane | grep -c ie
  25961
  # cat /usr/share/dict/british-english-insane | grep -c ei
  8255

But even at 3/4, not sure you can call that a good "rule", if it's wrong that often. But of course, that could be swung subjectively if the most used words follow the rule and the least used words break it... which I imagine is the case, since people do still use this rule, but I'm not sure of an easy way to argue the point.

Edit:

I suppose one way would be to assume that british-english-small (with 50083 words compared to insane's 638286) is comprised of the five thousand most used words (which I assume is roughly the logic?)...

Small shows that "i before e" is correct 88% of the time, and "cei not cie" is correct ~27%.

So in fact, limiting to most common words does make both percentages go up, the "except after c" is still much more often wrong than right, and the general "i before e" is... perhaps now high enough to call a good rule? Not sure.