I've had a similar issue. The sensor works fine when the MUX isn't installed, but it doesn't work with the i2c MUX. I think the issue is that the sensor needs a little time to calibrate and the bus is constantly switching out from under it since the while true main loop switches to bus 0 every second. That being said, I've tried duplicating the sample test code for the SI1145 inside GroveWeatherPI and disabled all the timers so that only sampleAndPrint is running and also disabled the 1 second reset to bus 0 in the main loop and I'm still getting 0s. Are you able to get it to work with the I2C Mux at all?
This is a very sensitive device and has gain variables of over 100 to one, so it looks kind of right. We have one down in Project Curacao that reads up to the limit (like yours does) in bright sunlight.