Sure, fairly easy to explain. Historically, favorability ratings have only a loose connection to actual voting behavior. For example, Donald Trump had consistently low favorability numbers during the 2016 campaign, and many analysts assumed this would prevent him from winning. He also currently has a 10-point unfavorability rating. Yet, he secured the presidency in 2016, and is currently even and trending up in electoral college projections, showing that favorability ratings don't predict and are often only loosely correlated with electoral outcomes. Similarly, current favorability ratings suggest Harris is struggling in terms of public perception and trending down, but those polls alone don't fully represent the complex dynamics of a presidential race. Note that your 538 visualization uses a two-year timeline, meaning you're missing a lot of granularity in favorability since she took over the candidacy. Since that initial jump when she took over for Biden, her numbers have been much more ambidirectional on that front: from low-to-mid 50s to high 40s, etc.

Looking at the current polling data, Vice President Harris is trailing Trump in several critical battleground states like Arizona and Georgia, with Trump holding a 5-point and 4-point lead in these states, respectively. While she holds a slim national lead in some aggregate polls, such as a 3.7% edge in national polling (less than Biden won by), this has not translated into a clear advantage in all key regions and, thereby, the electoral college. Importantly, Biden performed better in similar polls during the 2020 election campaign than Harris is currently performing against Trump.

Compared to Hillary Clinton at the same point in 2016, Harris is polling similarly or worse in some battlegrounds, i.e., on the whole worse. Clinton also struggled with unfavorable ratings but she managed to maintain stronger polling in many regions than Harris currently is up until late in the race.

The issue with relying solely on favorability ratings is that they only capture general sentiment toward a candidate and not the decisive factors like voter enthusiasm, turnout strategies, or key policy stances that can ultimately sway elections. They also don't capture voter intention, but rather how a broad sampling "feels" about a candidate at any given point in time. Moreover, favorability ratings can fluctuate based on short-term controversies or media coverage and don’t always reflect the strategic or demographic realities that decide elections. In short, while these ratings can provide insights, they shouldn't be overemphasized as predictors of electoral success, and they're certainly not useful as a counter-argument to the simple fact that Harris is doing worse than Clinton and Biden at similar points during their respective campaigns.

While it's frustrating and seemingly unbelievable given who she's running against, Harris is currently in serious danger of losing the election.