Looking at pages 50 & 52 and comparing them we can see that the number of responders are weighted down for FF, FG, SF, GP and IND/oth; but, weighted up for Labour. That is the adjustment and was probably done simply by scaling the responders of the over represented groups.
Something I’d forgotten to think about and which can be seen on the right hand side of those pages is that B&A also weight by likelihood to vote. That might go a long way to explaining why the ind/oths have not been uprated by a significant amount. As we can see the Ind/oths answered the likelihood to vote question with a higher amount of scepticism over their own desire to vote than the other responders. Personally I think that likelihood to vote filters are unnecessary given that fact that I’ve already mentioned that you’re already at a high likelihood to vote if you can be bothered to answer an opinion poll.