We have some wrong models but this does not mean we don't have biases.

The simplest way to disprove it:

If we did not have biases but wrong models, fixing models would make us unbiased.

But that works very rarely in real life.