Models help test test our understanding of past events. If you can create a model that reproduces past events, there is a chance that you have understood something about those events. Equally, however, there is a chance that you have created a model that only coincidentally works, because you forced it to do so.
When "scientific" articles write that a model "confirms" some conclusion, they are just wrong. Given enough free variables, you can construct a model that does anything you want. Models confirm nothing.
"The results confirm the North African Humid Periods occurred every 21,000 years"
No, they don't. Did the Sahara green every 21000 years? except when there were ice sheets? Only physical tests can confirm that. Do these greening phases correspond with orbital precession? Statistics can confirm that correlation. Running a model can test hypotheses of causality, but it doesn't "confirm" anything, least of all physical facts.