Matplotlib and numpy version conflicts not properly resolved

I am running a python script under the GMAT python plugin that uses numpy and matplotlib. GMAT 2020a is only compatible with Python 3.7. So looking at NEP029 the version of numpy that is compatible with Python 3.7 is numpy 1.15 or 1.16. At, the “Matplotlib min_dep_policy.html#list-of-dependency-min-versions” page claims the correct recipe is numpy 1.16 and Matplotlib 3.4. The obvious solution is to create a Python 3.7 environment for GMAT. However, the Anaconda navigator automatically installs numpy 1.20.5 and matplotlib 3.5.1 in the new python 3.7.9 environment.
When I execute the python plugin standalone in debug with the Anaconda chosen versions of numpy and matplotlib, I get the dreaded “multiarray” numpy import exception. There seems to be a problem with the Anaconda wheel. The Anaconda dependency analysis does not correctly identify the versions that are identified as compatible with Python 3.7 by the ecosystem providers, in fact the versions being installed don’t work.
I can go to conda and install the specific versions into the environment, but I am having enough trouble myself trying to figure out the combination of Python 3.7, numpy and matplotlib that are compatible. I had hoped that the science team at Anaconda had already figured these things out.
For the readers, the import numpy works in the python 3.9.7 base environment with matplotlib 3.4.3, and numpy 1.20.3. And also, using conda install, the numpy import also works with python 3.7.9, numpy=1.16 and matplotlib=3.4.