Optimum FL is dependent on the speed chosen. You can watch it change as you go selected or modify the CI. REC MAX is perfectly safe, if you know what you are doing, and I think there's only one key point to it:
Make sure you have the correct temperatures inserted for the predictions to be valid.
Other than that, provided you have modern day wind data, I'd watch for wind component difference between the two non-optimal levels. If equal, my choice is to take the higher one. Apart from what had been said already, FL340 vs. FL380 makes for a significant difference in choices available if you need to deviate around TSs.