No, the assumption isnt that its wrong, the assumption is the biblehash algorithm favors small rigs.  Remember the biblehash algorithm isnt written to reward the fastest proc, its written to reward who solves the biblehash chained verses with a solution- and that requires a full node- take a look at the code for the hash algorithm- it requires you to look at the block, get things every so often when the timestamp needs changed, chain the verses together and execute some business logic from the full node to hash it.

Now as far as how do we explain why the pool seems to favor distributed nodes over super machines, I have no clue.  Ive looked at the code for the solution reward, and its like this:
I hand out 20 equal easy problems to the team, and as you solve those equal easy problems you get a share.  Thats how it works.  Why multiple machines solve these equal weighted shares 'easier' is still beyond me.