The poll measures first startups, not all startups. Since successful founders have often started less successful ventures previously, the data makes more sense than you'd think.