eLife pursuing the niche of exclusiveness

Which I don't blame them really, they are better positioned than others to explore the limelight publications. I just read this FAQ about it:

Helping your findings achieve maximum reach and impact – Your article will be selected by notable life science colleagues from around the world and published alongside work that is judged to be amongst the most outstanding in its field. Impact statements from editors, along with plain language summaries and commentary from thought leaders will help to explain the significance of the work.  And, your article will be immediately and freely available to the world.

It's also mentioned the confidentiality of the editorial process, remuneration of academic editors, and it's emphasized their distinction from megajournals (PLoS ONE?), in that they are really into potential impact. My comments on twitter:

mixed feelings about @eLife_sciences. Looks perfect, except for emphasis on impact (glamour?) ow.ly/b0j2G
— Leonardo Martins (@leomrtns) May 18, 2012

some fields/subfields are just not attractive. And we are seeing a raise of poor works w/ high impact. good impact measures take time (1)
— Leonardo Martins (@leomrtns) May 18, 2012

and peer reviewers are not that good at evaluating impact and technical soundness at once. Editors might be better at evaluating impact (2)
— Leonardo Martins (@leomrtns) May 18, 2012

but still, real impact is a posteriori, At first we can mostly say if it sells -- and TV is here to show us how it ends. (3)
— Leonardo Martins (@leomrtns) May 18, 2012

OTOH I understand there is a niche to be filled by "next gen" publications, of more "exclusive" research (4)
— Leonardo Martins (@leomrtns) May 18, 2012

The level field of potentially significance is uneven, including its overlap with novel research -- a novel method might not be as fancy as an old method on a novel data set, or an old data set on a novel computing facility. Furthermore there is the problem of field glamour: an editor pressed to chose between two equivalent papers from fields A and B may consistently attribute more impact to field B. And we have been seeing that controversial research sells, non-replicable research sells, big names with loose oversight sells, and publication bias is a problem with devastating consequences for science. And corrections to flawed research do not sell, small niche development do not sell.
What I wanted to stress out but I was reaching my twitter-fu limits is that it's on "impact/significance" that reviewers can weight in with all weird subjectivities. So to help solving the increasing complains about peer review, publications should really reduce or eliminate reviewers' right to clairvoyance. In many cases reviewers can really spot a game-changing manuscript or detect a dull result. But in many other cases they will even unconsciously see more platitudes than there actually are, or praise a work because of its implications neglecting its truthfulness, or even mistake a poor phrasing for an uninteresting paper. And the secrecy of the review process encourages such behavior. I'm not saying to get rid of impact projections, but please don't leave it to reviewers -- the fact that academic editors are being paid can be a good justification for them assuming this responsability (maybe this is their plan anyway.)
There's much more to be said, lins to be linked, but my coffee break is over -- and I might be missing relevant information about it. In the end I really like the idea of eLife, although it seems to be taking a very conservative approach -- perfectly understandable given the risk aversion of stablished brands, but a shame to see one of the best recent opportunities lost (e.g. to offer post-publication impact evaluation,  to accredit peer reviewers by publishing their reviews, allow updates on article, and so many other ideas that in my naiveté I thought scientists might be willing to explore.)