Hi Oscar,
I also wanted to bootstrap data using Mathematica and to be confidant about my results. Once I was happy about the methodology, I was able to get good agreement with test results I had obtained using Excel.
As a further test I tried replicating the results referenced in your post 'Make a difference: the alternative for p-values'. I managed to get good agreement with the Bootstrapped Medians, the differences between the Medians and the Confidence Intervals.
As far as I can see, the reason for you not not getting agreement and for your charts looking very different is that instead of Bootstrapping the Medians you have Bootstrapped Skewness.
Ronald Smith