Using the formula you would arrive at 5Rs. Pay attention to the spacing given in the formula and the spacing of your actual carriers. The 4-orthogonal FSK system is not at the minimum frequency separation; you could actually fit in 3 more carriers that would still be orthogonal and thus have the same bandwidth as a 7-orthogonal FSK system.
That extra 1 and 0 could come from different signals around 431MHz. Since you are not doing any lowpass filtering you'r demodulator gets whole 400kHz band (your USRP sample rate). I'd suggest putting lowpass filter after USRP source and double check signal you are receiving with QT GUI frequency sink.
Okay, I figured it out myself. I will share my solution. After some googling I realized that this is clock recovery problem, so I used Clock Recovery MM block (with default parameters except for Omega). From Clock Recovery MM
If the input signal pulses do not have peaks (e.g. rectangular pulses) the input signal should be conditioned with a matched pulse ...