Issues where people disagree are likely to be mixed issues, where making changes will do harm as well as benefit. That is exactly why people disagree. So working on those issues will tend to do less benefit than working on the issues everyone agrees on, which are likely to be much less mixed.
A disagreement could resolve into one side being mostly right and another mostly wrong, so actual harm+benefit isn't necessary, only expected harm+benefit. All else equal, harm+benefit is worse than pure benefit, but usually there are other relevant distinctions, so that the effect of a harm+benefit cause could overwhelm available pure benefit causes.
The disagreements I was talking about - which I clam are many, perhaps most, disagreements - are not about unknown or disputed facts, but about conflicting values and goals. Such disagreements can't be resolved into sides being objectively right or wrong (unless you're a moral realist). If you side with one of the sides, that's the same as saying their desires are 'right' to you, and implementing their desires usually (in most moral theories in practice) outweighs the cost of the moral outrage suffered by those who disagree. (E.g., I would want to free one slave even if it made a million slave-owners really angry, very slightly increasing the incidence of heart attacks and costing more QALYs in aggregate than the one slave gained.)
The article is here.
The book is by William MacAskill, founder of 80000 Hours and Giving What We Can. Excerpt: