Making the Biden/Trump gap larger wouldn't necessarily make them less correct if they reported a large percent of undecideds instead. I agree it looks bad, but it has to be studied on it's merits
There's more to a poll than just a number for each candidate. Look at how they got to their numbers and compare it to how that demographic actually voted and you will see they were majorly wrong. For example FL Cubans +55 above.
Yet we should all disavow the partisan republican polls despite underestimating Trump, while the extremely unbiased MSM pollsters that were off by 5-15 points must be explained by the 3% undecideds.