good summarization , but some problems from my point of view :
1. due to gausi filiter , GMSK signal is not so smoothly consecutive and still exist some deviation , that we can see some offset in constellation
2. for Gsm, TX and RX are time division , may some TX will get into RX channel , but how it is affect LO of TX modulation . it's the thing inthe chip , do MTK give some design consideration , in my opinion , isolation should be concerned during chip design , and the possible affected and unwanted interference should be defined in chip application level.
3.should give a clear interpretation difference between modulation and up conversion , maybe some confusion here by some ones
As for TX get back to TRX from RX path, it is normally attenuated by RX SAW filter. The worst case is qual-band or Tri-band design, GSM1900 TX get back from the DCS1800 RX path, due to they share ~30M freq range and the RX SAW filter cannot provide a good attenuation.
Agree with Ning about his point 1&3, expecially for 3, It make me confuse for quite a long time.