The title of the article was so misleading they actually had to change it.
Correction: The original title of this article incorrectly suggested that Neil Ferguson stated his initial model was wrong. The article has been revised to make clear that he provided a downgraded projection given the new data and current mitigation steps.
1. He was wrong about the rate of reproduction, which he now estimates around 3 rather than 2.5. This means there are
more infected people. Since there are more infected people the mortality rate automatically decreases. But we're talking about a reduction in relative terms, not a reduction in absolute terms if lock down measures were not taken.
2. The revision takes into account the lock down. That's why the number of deaths plummets.
PD: It's worth to point out that the criticism to the model by Oxford Professor states that the high incidence will trigger herd inmunity sooner, and thus lead to fewer deaths.