Rayleigh Channel power normalization

I have to use a Raleign channel as built-in function. And there some misunderstanding.

I set zero-delay and zero-gain for that channel. So, I expect the channel witouh fading, because all parameters is zero. Code here:

clear all
close all
clc

Fs = 72e6;
M = 4;
dpskMod = comm.DPSKModulator('ModulationOrder', M);
dpskDemod = comm.DPSKDemodulator('ModulationOrder', M);
numBits = 1000; % Each trial uses 10000 bits.
numTrials = 10;  % Number of BER computations

PathDelays = [0]*1e-9;
PathGains = [0];

rayleighCh = comm.RayleighChannel(...
    'SampleRate', Fs, ...
    'PathDelays', PathDelays, ...
    'AveragePathGains', PathGains, ...
    'NormalizePathGains', true, ...
    'MaximumDopplerShift', 0, ...
    'PathGainsOutputPort', true);

nErrors = zeros(1,numTrials);

for n = 1:numTrials
   tx = randi([0 M-1],numBits,1);
   dpskSig = dpskMod(tx);
   [fadedSig, pg] = rayleighCh(dpskSig);
end
disp(pg(1));

But, each run I have different results. For example:

-0.1019 - 0.0337i
1.4275 + 0.7458i

It's means that signal's gain will changing each run. Why?

With delay==0 and gain==0 I expect constant channel for every run.