First time, the DIO0 pin was shorted with Vcc.
This time with the changed circuit (reverse orientation), the MOSI and MISO pins are shorted to GND.
I have checked the connections again and again for any shorts or dry solder points but have found none. But if you insist then I can make another board and see if I get the same problem. Otherwise, I only have one option but to replace the MCP1702 regulator (which I don't have at the moment and is a bit difficult to get).
Another piece of info is that the circuit is taking around 150mA of current when transmitting (for a couple of hundred milli-seconds) and 30mA when the receive window (with a timeout of 20sec) is open. But if the current limit of the regulator was exceeded then the Arduino should reset or something, the RFM95W should not burn out IMO.