Attenuation Constant for waveguide

Hi,

I try simulate rib waveguide with silicon as a substrate, lower cladding silica, core glass and upper cladding air in order to find attenuation constant with respect to the lower cladding thickness. However, when I try simulate it, it seems that attenuation constant become lower as lower cladding is thinner. This is not true. Can anyone advice me, which step that trigger this problem?

Thank you