LPF load proportionality factor riks random

Hello,

My model is simply supported I-shape S4R beam. 

Could anyone explain me, why Static Riks method with an *imperfection returns different/random results for the same analysis runned two and more times?

Let's say that I set maxium load proportionality factor = 160.

For the first time, the solver needed  13 increments to exceed the limited value.(LPF=162.69)

For the second time, the solver needed 23 increments to exceed the limited value.(LPF=164.403)

How can i get correct results?