Your models seems to report confidences that are not evenly distributed between 0 and 1 but have an upper or lower boundary. This is not a problem per se but a bit unusual.
Also, the predictions reaching this boundary seem to be all false. This is why the red curve goes to the right, not upwards.
The model could still use some improvement, and you should check the process for operations that change or limit the confidences.
Answers
Your models seems to report confidences that are not evenly distributed between 0 and 1 but have an upper or lower boundary. This is not a problem per se but a bit unusual.
Also, the predictions reaching this boundary seem to be all false. This is why the red curve goes to the right, not upwards.
The model could still use some improvement, and you should check the process for operations that change or limit the confidences.
Regards,
Balázs