My guess is that there are two major factors that can account for most of the difference:
1) DxO reports '35mm-equivalent' lp/mm numbers, which are 2 times smaller than the real lp/mm numbers, due to the smaller size of the 4/3 sensor.
2) DxO appears to compute the resolution figures from the individual raw R, G and B channels, whereas Lenstip uses a demosaiced image. It seems reasonable that the Lenstip numbers exceed those from DxOmark, because information from all channels is used simultaneously.
