Interesting that there's such a degree of variation.
Looking at it from a practical ( non-radon ) point of view; assuming all other things are equal -
The duty cycle of the instruments and therefore the current gain must be significantly higher in the classroom environment. Unless the levels of humidity are condensing, the variation is too significant within the claimed similar environmental conditions.
However, the noise/speech cycle could be hugely different, especially with higher ambient noise like PC fans or AirCon units. This is in addition to the actual increase in speech in class and the length of use as pointed out above.
If I had to nail it down to one reason though, it would be the feedback manager - having to constantly invert and squish a signal is going to drain more battery than virtually anything else. I had something similar with a client who used to get way less usage when wearing a hat. Do you sit in a study booth when doing your work? Is there a constant high pitch noise in your class environment?
I'll have a beer on the FBM over anything else.