Yeah.. well there is that..
It's simple - large chunks of the population do not actually trust people in authority. We may not all agree on which ones are untrustworthy but it would be next to impossible for everyone to agree on someone we
all trust. You trust the civil service. That's fair enough. But there are numbers of people who don't. Who can site... for example, it was civil servants who went to nice lunches with Vodafone executives then decided six billion pounds worth of taxes weren't actually worth collecting after all.
You can find similar examples of apparently untrustworthy behaviour in pretty much every branch of the state - police, judiciary, civil service, politics... as soon as you invest authority in someone you leave them open to at least accusations of abuse of that authority. The only way to ensure that isn't the case is to give the authority to everyone. If you do that, the authority itself has no value and therefore isn't worth compromising.
That's the point of publication. If everyone has access to the information, nothing can be hidden and you don't need to trust anyone else to decide what you can or cannot see and what is or is not relevant. I think retrospective publication is enough in this regard to act as a deterrent to misdeeds, at least in terms of casual wrongdoing. No overwatch scheme is infallible, so lets have the one that does most good for least effort.