I don't exactly understand, what you have been simulating. Particularly, I don't see how yout set up a gain < 1 for a Sallen Key filter.
As said, a Sallen Key butterworth filter can be designed for different gains (Af values). The RC dimensioning has to be adjusted respectively, because you get an additional term (1-Af)R1C1s in the transfer function's denominator (besides the numerator factor Af).
The gain value of 1.586 will be obtained by adjusting Af of a filter with equal R and C to match the butterworth transfer function.
You can calculate the required gain value for a specific pole pair quality factor for the equal R and C case. It's
Af = 3 - 1/Q
Because Q is 0.71 for the second order butterworth, you get the said Af value.
The transfer functions is however only valid for an ideal OP, if you put in a real one with finite gain and non-zero output impedance, you get deviating results. The curves shown in your last post are apparently simulation results with a real OP. Generally, increasing the filter gain will reduce the loop gain and result in a higher OP related deviations.