Negative resistance oscillators that designed wrong

I came to a conclusion that authors of some papers on negative resistance oscillators (maybe most of them) does not do extensive stability analysis. I just tried one layout in simulator few days ago, and it gives a wrong frequency. Analyzed few other designs just by S-parameters. I did it because started to become crazy trying to optimize my oscillator for instability, i kind of making a big spreadsheet for negative resistance oscillators, and there are plenty of formulas, and many plots. The good thing is i learned many things. Here are details:

I just made simple formula to calculate distance between points in polar coordinates. Idea is that i calculate μ1 and μ2 stability factors not for center of Smith chart, but for arbitrary loads. And i did all that positive/negative sign correction for different cases of stability. So i build all those plots for my "alternative" μ-factors and phase plots (reflection coefficient required vs real). Idea is to show oscillation conditions graphically. And what i see

1 - many dips on μ-plots, and they are negative. It means, that reflection coefficients at unwanted frequencies are reside inside unstable regions! Yes, there is a dip at required frequency of oscillation, but how about all others?

2 - ok, someone can say that phase is not matched. But surprisingly it is matched (required phase of reflection coefficient and Гin have opposite signs, or it is can be seen as inverse,etc.etc many ways all rechecked thousand times). In best cases oscillation condition matched at two frequencies.

Is not it a problem?

So i see one proper way to make negative resistance oscillator is to start by choosing few active devices, and compare theirs instability regions. Then somehow best can be chosen, where are no so much "wideband unstability". I mean all those dips and phase matches.

Another way is brute-force approach. I think that just 3d or 4d multidimensional plots must be drawn. I mean each coordinate is a reflection coefficient, and cell value is μ-factors, or phase difference, magnitudes of S11 S22 etc.. So after such "heat maps" are built, best candidates for reflection coefficients are chosen.

Deeper i go - the worse!
Wideband. Authors admit in few papers, that they become not so wideband in reality because of some mysterious parasitics. I doubt it. It is more likely that particular design is not analyzed for instability. I am almost sure, that it is not some mysterious parasitics, it is the ones that reside in S-parameters. And they will give a huge dip on alternative μ-factor plot at wrong frequency of oscillation. Their oscillator "jumps" to that frequency, when varactor gives reflection coefficient. Phase matches at both frequencies, required and wrong one. Why it jumps - who knows. So it works as it works. Maybe it will jump someday some other way. And it looks like empirical voodoo tuning.

Also. DRO. Series DRO, when optimized for instability at required frequency (good negative resistance) appears to have hidden oscillator! Not always but often. Even with 50 Ohm termination. And that is because at some unwanted frequencies instability optimization moved unstable region right to 50-ohm. Not exact match, but pretty near.

Does not want to offend any authors, just really becoming crazy of all that calculating stuff.
Please comment on this issue.

I have a hint for you: many papers say things should work, but they lie.