Iterative Phase Estimation with noise vs standard Quantum Phase Estimation with noise

I am doing Qiskit Lab 4 about Iterative Phase Estimation. I created a circuit implementing IPE for theta = 1/3 (phase of 2pi/3). Here’s the circuit:

It seems to do okay if I run it without noise in a QASM sim:

Then the lab compares how IPE performs compared to QPE when you add noise – in this lab they use NoiseModel.from_backend() from the Athens IBM computer to simulate the noise.

The lab guide says I’m supposed to get something like this:

In reality I get this:

I have zero idea how to approach this problem because I tried inputing several values of theta to simulate in QASM without noise and everything seems to be fine there. Any ideas?

For reference, the code implementing IPE is below:

import math
theta = 2/3*math.pi
IPE = QuantumCircuit(2)
cregs = []
IPE.x(1)
for i in range(n):
    cregs.append(ClassicalRegister(1))
    IPE.add_register(cregs[i])
for i in range(n):
    IPE.reset(0)
    IPE.h(0)
    IPE.crz(theta*(2**(n-i)),0,1)
    if i > 0:
        for j in range(i):
            IPE.rz(-math.pi*2**(j-i),0).c_if(cregs[j],1)
    IPE.h(0)
    IPE.measure(0,cregs[i])

UPDATE: Apparently specifying the initial_layout during transpiling is crucial as it seems to have worked!