I am definitely saying its a bad thing - at least ethically. Like I said, the disheartening thing is unless the system changes, its probably good for your career. For the record, I should have said "successful early career folks" (i.e. other grad students and young faculty). Have yet to have a senior person say that, though I'm sure many do.
And yes, it basically boils down to the results looking better that way and the relative ambiguity about how to handle data. I also know several people who have admitted they check for outliers only when hypotheses don't come out...not when they do. I'm sure the prevalence of that is just astronomical.