Of course the model accounts for that. And in the past, the first few contests are extremely predictive of the final outcome. You’re totally and obviously wrong.
Iowa and New Hampshire are so different than the other contests. Most states only see a week or so of campaigning and have to make decisions without ever seeing the candidates.
Iowa has months and months of intense campaigning. For the most part they feel if they just pick the front runner why did they go through that? They like to be contrarian and humble the front runner.
Then when New Hampshire votes they don’t want to merely ratify what Iowa does so they vote for someone else.
Then the real race starts and the front runner shows why they were the front runner and win easily. Name recognition is usually the decisive factor.
It’s easier to regain the support you lost than gain support that you never had. People disposed to vote for Biden were disillusioned by his early performance and found other candidates, then flocked back to Joe when they saw him gain his footing. It was indeed a historic comeback but only because disproportionate weight was given to the first two contests with lily white populations and the third was a caucus where a minority of passionate supporters wins the day.