Meeting oscillation conditions at input or output port gives different matching?

I am doing calculations on negative feedback oscillator. Here is some confusing part. If oscillation condition is met either at input or output port, then circuit is oscillating. I understand that. But when i calculate reflection coefficients for both cases, i get different matching networks. I thought that formula of Gout and Gin for both cases will provide exact results.

I mean:
(case 1)
if i choose some Gsource near to Gsource=1/S11 and
Gout=S22+(S12*S21)*Gsource/(1-S11*Gsource)
Gload=1/Gout


(case 2, oscillation condition at output port)
if i choose some Gload near to Gload=1/S22 and
Gin=S11+(S12*S21)*Gload/(1-S22*Gload)
Gsource2=1/Gin
then i get Gsource2 that is pretty far away from Gsource in the case 1.

For example, in first case i choosed
Gsource=PolarToRect(0.35,-170);

In the second configuration Gsource is not choosen, it is result of calculations:
Abs(Gsource)
Ans = 0.014390937
Arg(Gsource)
Ans = -179.931423526

My idea is that Gsource2 must be more similar to Gsource because we use
these formulas:
Gout=S22+(S12*S21)*Gsource/(1-S11*Gsource)
Gin=S11+(S12*S21)*Gload/(1-S22*Gload)

Please comment. My calculations somewhere wrong, or it is true that we get two different configurations in these two cases?

In other words, i thought that choosing Gsource=1/S11 will automatically give Gload=1/S22 through using Gout=S22+(S12*S21)*Gsource/(1-S11*Gsource)