Hi Mike,
The two numbers define the range of pixel values in the integrated image, that is, the minimum and maximum pixel values. As the resulting image must be in the [0,1] range, it must be either truncated or rescaled. By default, ImageIntegration rescales the result in these cases, which preserves the entire integrated data.
You can disable this behavior by enabling the truncate on out-of-range option. This will clip negative values to zero, but will output two images without any scaling and referred to the same numeric range.