Hello, I'm having trouble using the libtea library's "void libtea_calibrate_flush_reload(libtea_instance* instance)" function

[dnffabs]

There is a significant difference between the threshold computed by the function on the image and the threshold computed by the libtea library "void libtea_calibrate_flush_reload(libtea_instance* instance)" I don't know why, but the threshold calculated using the function on the image only fits the logic of my experiment