The quote you cited says "no evidence the saliency model explicitly encodes male gaze" emphasis mine. Your re-phrasing however changes the meaning in your question, "historical biases not inherent in the model".

Explicitly encodes != inherent in. The difference between the two might answer your question. You can solve for something that is inherent in a system (ie, this), but was not explicitly encoded. They are trying to say it isn't their fault specifically, they find no evidence this was done on purpose or in a targeted way, but that it happens as an accident / lack of testing on the current model.