The whole setup to run something like this already runs in the high thousands. If you plan on doing anything worthwhile (legally) you will be spending nearly $5,000 on a benchtop RF cage. If you plan on needing amplifiers to get more than 10 feet of range you will be spending $2,000 on an amplifier. There are no shortcuts in RF, operating an illegal basestation can very easily affect E911 service and put you in jail for a very long time if your basestation stops someone from receiving emergency services.
This group of people are not the developers of the open-source GSM basestation stack they demod, funding them does not contribute to the OpenBTS. Fairwaves built a business trying to commercialize the readily available OpenBTS software stack, which was developed by Range Networks. Range split and the original author of OpenBTS, is the project maintainer behind YateBTS, which is a totally revamped version of OpenBTS with support in the works for LTE. So if you care about GSM, and LTE basestations go directly to the source http://yatebts.com
They are not using industry standard terms in the datasheet for some reason. I believe the 160MHz is the analog bandwidth of the device. There is no real-time bandwidth (RTBW/IBW) specification, which is the number the defines the widest channel bandwidth of a device. It is likely that the 60MSPS bandwidth can be shifted within that 160MHz analog bandwidth using the on-chip NCO to hop around.
Throwaway to avoid giving too much away but I have used the AD9364 in a few point to point links at my current employer. Its ADC sampling rate is 640MSPS [1]. This oversampling is important, if not outright necessary for noise shaping, and appropriate baseband filtering. I am not quite sure how the LMS7002 chip can achieve a usable 56MHz channel bandwidth with only 61MSPS.
Anyway, I'm not quite sure what ESA will be able to do that Canonical couldn't. This whole app store idea for radio seems very farfetched.