You are pointing at evidence that 538 correctly called 11/12 races using statistics, and their confident call on a Biden president withstood a 4-7% swing (!!).

The existence of bias doesn't invalidate their predictions. Everyone knows that polls can be badly off target in a biased way - that isn't a new phenomenon.

When they talk about X% chance of Y being president they should be optimising to the outcome, not the margins.