The El Niño-Southern Oscillation (ENSO) is the primary source of seasonal predictability over North America, particularly in winter, and so there is considerable scientific and public interest in the strength of the winter El Niño or La Niña. Despite that our state-of-the-art computer forecast models were forecasting only a marginal La Niña for the winter of 2024/25, those same models predicted winter impacts over North America that were consistent with a moderate-to-strong La Niña. In contrast, those models correctly predicted a strong El Niño in the previous winter, but the predicted impacts over North America were only consistent with a weak El Niño. We determine that the main reasons for this apparent mismatch over these two winters were (1) exceptional warmth in the deep tropics outside of the equatorial Pacific that muted the North American signal in 2023/24 but strengthened it in 2024/25, and (2) a separate pattern in 2023/24 tied to sea surface temperatures outside of the tropical Pacific that destructively interfered with the El Niño signal in the winter of 2023/24. These findings highlight challenges in monitoring ENSO in a rapidly warming world and also point to other sources of winter seasonal predictability.