The problem is the Guardian's write up of the poll, which is an impressive effort in trying to make something out of nothing.
I know journalists have to write stories and proprietors sell newspapers and stuff; and I know "Poll shows no real changes" is a boring headline. But a 1% change in the number of people who think Cameron is doing a good job doesn't indicate anything, other than that the polling company's methodology is robust enough to produce broadly similar results.
Does the Observer not understand the concept of statistically insignificant variation? If Labour's support has dropped by 3 points, it probably ins't duse ot anything that has happened; it just tells us the last poll was probably a bit too generous, and this one a bit too harsh.
Pretending a change from one poll to another, and reading Grave Import into it, is pointless at best, and at worst a rubbishy attempt to set the agenda or engineer a crisis. So they get to write more stories and sell more newspapers.