It's true, but the review process (while not the publisher) does add value. It's really useful to highlight the gaps in an authors work and suggest additional analysis they could do to support the work and make it more useful. Or even highlight areas where the work is not reproducible (algorithms not sufficiently well described, data not available etc).