The Telosb has a 12 bit adc which means you would expect a range of values from 0 - 4095. The values for I and Q in the above 4 data points do seem to be in the range of values generated by the bumblebee when 0A is treated as the most significant byte and B3 is treated as the least significant byte. I would assume therefore, you are doing the right steps (0x0A << 8 + B3) to process the data. Can you please provide us with a simple plot of sample value (after the conversion) vs sample number instead of the spectrogram and possibly more details on the environment the test was performed in like indoor vs outdoor, other objects in the scene and height of the radar from the ground ?
Note that the radar also "sees" behind and to the sides albeit for shorter distances, but please keep this in mind when describing the environment.