I don't quite understand the significance of removing the worst team from the equation. It's perfectly possible for the worst team to get a decent result - for example, England beat the second worst team but only drew with the worst. It doesn't strike me as an accurate way of assessing group difficulty.
I think KG's SPI method is best.
Well for Brazil's group using the mean makes the group look relatively easy, due to North Korea's lowly ranking. Using the median gives a more realistic assessment of the strength of the group, as teams need to finish in the top 2 the median value gives a theoretical ranking figure you need to exceed to qualify for the next round. All this being said none of these methods are ideal, all forms of statistical analyses have their strengths and weaknesses. I was just trying to add another way of doing things to the mix.