It's a weakly predictive indicator. Yes, it tells you something, but not much. Extrapolate at your peril.

Look at it this way: Most of the people interviewing you are other programmers – the exact opposite of skilled interviewers. Predicting the day-to-day reality of working with them based on their default interviewing style is just as unreliable as them judging you based on your performance at solving silly brain teasers under stress.

Despite how screwed up this all is, most people are going to end up in many interviews throughout their lives, probably on both sides of the table. So it's best to realize up front that the situation is screwed up but changeable so that when you end up in an interview you'll know to ask for changes.

Yes, it's still an indicator. But so is your performance in the interview, correct? But your argument is that your performance in that style of interview is a weak indicator of your actual performance. Then I think it's also fair to grant that a company starting out with the standard interview style is a weak indicator of what it's like to worth there.