I use optimize(56, 8, 112, 8) with additional smoothing period * 2. You are correct, it has a little effect on individual component and no significant effect on overall portfolio performance (-1% AR) which I consider a statistical fluctuation. But it has an effect nonetheless. What actually happens is that new backtest with new MMI changed OptimalF of one component. That changed reinvesting margin resulting in a visibly different portfolio statistics.