COSMOS fitting seems to depend on the number of channels used for processing

[jc-brandeis]

COSMOS fitting on a set of 2-channel data gives different results when the 2 channels are processed separately or together:
spot detection in channel 1 shows very low p_specific for all AOIs in all frames when processed together with channel 2.
spot detection in channel 1 shows reasonable p_specific when processed by itself.

[ordabayevy]

A shared parameter between two channels when analyzed together is proximity. I wonder if the second channel influences the value of proximity parameter which in turn affects the p_specific for the first channel.