The way I understood it (by way of the paper's gold panning illustration) is that with a progressive tax system based on total aggregate income (i.e. you don't distinguish rent income from non-rent income), high marginal tax rates result in a multitude of rent-seekers each collecting a fraction of the total available rents. This is so because as their income for the year accumulates and their marginal tax rate increases, the tax rate overcomes the collection effort required. Once they've maximized their income they rest for the remainder of the year. If they're not collecting rents that means someone else can. The market in rent-seeking supports a larger number of rent-seekers, which diminishes the base of people doing productive work.

By contrast, imagine if all the rents went to a single rent-seeker. In that situation only one individual is doing unproductive work instead of many. Assuming that the rents were going to be collected regardless, then we're better off if there's only one individual pursing that unproductive activity than many.

How do you ensure there's only one rent-seeker? By creating conditions that favor someone monopolizing the rent-seeking market. You can do so by taxing high income earners less than you might otherwise, preserving the incentive for a single rent-seeker to collect all the rents. The lone rent-seeker is taxed less, but that's okay because he's keeping everybody else out of the rent-seeking market. Taxes on other high-income earners is less but that's okay, too, because they're doing productive activity--generating wealth and thus more taxable surplus--and now there's more of them.

Like the Laffer Curve the devil is in the details.[1] But it's an interesting point.

[1] I don't mean to impugn the paper by comparing it with the Laffer Curve and the toxic political economics surrounding it. I just mean to say that the quantitative values matter.