How to change floating point model

Thank you for your answer! Yesterday I looked into the config options as well and I suspect that you are right and that I need to change the UBT code. For testing purposes I completely removed the

if (!WindowsPlatform.bCompileWithICL)
{
    // Relaxes floating point precision semantics to allow more optimization.
    Arguments.Add("/fp:fast");
}

part and after recompiling the engine it seems as if the bug has gone. However, the result is still slightly different from when run on a standard vc project configuration, so there might be another difference in the compiler arguments or a bug in my code.