What is phase delay through tapered microstrip line? (simulation vs reality)

Hello all!
My current design requires use of T-power dividers and very phase sensitive.
But for simplicity I ask a simpler case, but with the same problem. Just a straight microstrip line with two two quarterwave tapers.

Imagine we have a tapered microstrip line, total length is one wavelength. This line consist of four quarterwave sections:
[port1]-[quarterwave 50 Ohm]-[ quarterwave X Ohm ]-[ quarterwave Y Ohm]-[ quarterwave 50 Ohm]-[port2]

The problem is S21 angle of this simple line is varies from 0 degrees to around -20 degrees (fdtd, rf simulator with spice support, etc.).
I expect it to be near 0 degrees. Because I have many such sections in my design, -20 and even - 10 degrees shift is unnacceptable.

How to solve this problem before actual fabrication?

Unclear question. What are you varying? Frequency or other parameters? What's the frequency range? How are you measuring phase? With port extension according to the line length or without?

Did you consider the effect of effective dielectric constant on the wavelength i.e. guided wavelength? Simulators may not simulate the materials perfectly.
Examine its sensitivity, introduce an error of 10% on the dielectric constant and calculate what it causes on the amount of phase shift. 10 degree of shift is not looking high to me.

FvM, i use constant frequency value.
The only thing I change is line configuration:
4 quarterwave 50-ohm sections in series
1 quarterwave 50-ohm + 1-Y ohm + 1-X ohm + 1 quarterwave 50-ohm sections in series
X,Y around 30...40 ohm.
Each time it is a full-wavelength line, also I put them in repeating series.

I would expect S21 0 degree total phase and varying magnitude caused by different standing wave patterns. If the phase is different, the electrical length of each line is probably not λ/4.