MLX90614 sensors with restricted fields of view

I have three MLX90614 sensors. Two of them are yours -- I2C devices with 35 deg and 5 deg fields of view. The third sensor, from another vendor, does not have a tube that restricts its physical field of view. Each of these sensors give roughly comparable (very cold) values when pointed straight up at the open sky. Of the three, your 35 deg sensor seems to be the least influenced by ambient temperatures on the ground. How do your two sensors, both with tubes extending over the sensor housing, ignore the temperature of the tube, which it seems to me the sensor should "see" as part of its field of view.
I note that when I placed my own tube around an MLX90614 sensor and pointed that device at the open sky, the results very clearly indicated that the reported temperature was some average of the ambient air temperature around the tube and the very cold open sky "effective temperature."
This is certainly NOT a criticism of your restricted FOV sensors! Rather it is an attempt to understand why they work as well as they do. For my purposes, your 35 deg FOV sensor costs only a couple of dollars more than a "bare" sensor from other vendors, so it will be my first choice for these kinds of open sky measurements.