The loader code looks pretty straightforward. For Win32, the overlay signature ("112358" in a binary I looked at) is followed by:
dword codesize; char native_code[codesize];
dword datasize; char init_data[datasize];
dword importsize; char imports[importsize];
dword symbolssize; char symbols[symbolssize];
dword ressize; char resources[ressize];
dword optsize; char options[optsize];
After fixing up imports and relocations the loader jumps to the start of loaded code.
Don't have time to look further but it doesn't look too complex :)
Note: Registration is required to post to the forums.