By default matching is done in a completely case insensitive way. This makes sense for the protocol and host components which are defined as insensitive by RFCs however it does not make sense for path.
A byproduct of this is that when using request history the values for path, qs etc are all lowercased as this was what was used to do the matching.
To work around this when building an Adapter or Mocker you do
with requests_mock.mock(case_sensitive=True) as m: ...
or you can override the default globally by
requests_mock.mock.case_sensitive = True
It is recommended to run the global fix as it is intended that case sensitivity will become the default in future releases.
Note that even with case_sensitive enabled the protocol and netloc of a mock are still matched in a case insensitive way.