In the past, these types of surveys are filled with methodology errors.

A Seattle based survey I see regurgitated often simply asked where their LAST place of residence before becoming homeless was, and it included stays in shelters as residence... So essentially anyone who had been there more than a few months was counted.