well another issue is that ML uses 64 bit kexts only and the NVHal40 kext is only 32 bit so there is also that factor. also i am not sure lion kernel will work on ML i know the ML DP1 kernel still had support for 32 bit kexts but it will bitch alot while booting in verbose. you would have to see all the drivers being used in Lion for graphics and add them in to ML as well. but it still wont have hardware acceleration. i tried on a similar model the d620 nvidia and it was very unusable compared to snow or lion when it worked correctly.