That's not necessarily survivorship bias. Survivorship bias would be "Here's 10 successful companies, do what they do". If you analyze a pool of random companies and then consider what the successful companies do differently then that would be decent scientific method (though you should form hypotheses first, really).

For example, if "risk-takers" both sometimes do very well and also often fail then you expect to capture the failed risk-takers in your initial sample, so you're not just biased by the successful risk-takers (said survivorship bias).

I'm not familiar with the book, but the snippet you reference describes something much closer to proper scientific method than most of the examples of this phenomena.