For the 4-switch buck-boost converter, if designed well, efficiency will exceed the SEPIC configured device. For 50W load a controller and external FETs would be best, and provide low losses in either buck or buck-boost/boost modes. The SEPIC may have fewer active components with a switch and diode, but requires two inductors (or coupled inductor) yielding added losses, and also the high RMS current in the coupling capacitor adds to the losses.
Noise and EMI can be comparable, although the hard switching of the capacitor in the SEPIC at peak-peak voltage levels that are essentially Vin+Vout, EMI can be much higher. Softer switching of the FET and diode can reduce that. The lower switching voltages of the buck-boost circuit, in either buck or boost modes produces less EMI with a good layout. In the region of Vin close to Vout, the multiple switching cycles introduced in buck-boost mode can add noise, although can be filtered since it will be higher frequency.
Both SEPIC and buck-boost topologies can be compensated for stable operation. The SEPIC will usually have a lower frequency rolloff due to the capacitor and inductor in the power output loop, requiring compensation bandwidth to be lower, below 10kHz is common. The buck-boost can be designed for stable operation with bandwidth approaching 20khz over the input voltage range based on the single inductor and output capacitor, with both boost and buck modes being stable.
MPS has many buck-boost products, including the MPQ4230 with 6A switches that can handle 50W loads.