According to toms hardware, that firmware update drops the (under cpu load) power usage from 7.9w to 7.6w. That sort of difference will likely be difficult to separate from the noise of a non-synthetic load.
The IR images were taken at idle, so naturally the relative differences look larger, as the USB controller is a much higher proportion of the used power than when the SoC is under load. Aside from IR cameras being a /terrible/ way of measuring heat generation or cooling performance - as it makes small differences look large, well within the range of error due to different air temperatures or airflow or different stages of heat soak.
The IR images were taken at idle, so naturally the relative differences look larger, as the USB controller is a much higher proportion of the used power than when the SoC is under load. Aside from IR cameras being a /terrible/ way of measuring heat generation or cooling performance - as it makes small differences look large, well within the range of error due to different air temperatures or airflow or different stages of heat soak.