New analysis from Kawanomoto-san dot.pdf quoting him :

I tried to calculate flux attenuation around a dot by using ray tracing non-telecentricity of WFC brings alpha dependency if the center of fiber circular movement meets with the center of a dot. But, the sign of alpha dependency (curvature) is opposite @r=200mm. Small offset which is depends on field radius (linear, quadratic, and cubic - linear is some of scale change, cubic is distortion error, quadratic is not natural in optical calculation) may explain observed flux attenuation around a dot, but I can not find exact mechanism of alpha-dependency.