NavList:

Discarding it has much the same practical effect as retaining it, if the problem causing it to appear as a "real misfit" can be resolved and corrected.

The unsafe procedure, it would appear, is to accept uncritically data points (sights); to treat them as all equally valid, since as has been pointed out, one "real misfit" can significantly skew otherwise sound results.  This is a crucial difference between use of the slope technique and averaging, with which it tends to be erroneously confused. 

On another topic George has admitted recently:

"A few minutes ago, I posted the following, in which there was a silly error.
It doesn't (I think, at first sight) affect the conclusions, but that will
call for a bit more thought."

I wonder if George might also consider whether his earlier somewhat glib conclusions as to this matter could also benefit from a: "bit more thought" ?