Nonsense. I sincerely hope Paradox has done a more sophisticated analysis than that!
Generally, anything that expands distribution is good for business, as long as it's not actually money losing. It depends on a number of things: what are the actual QA costs, relative to that 5% of sales; are those cost truly variable or are some fixed; how much cost savings would you actually realize by closing a distribution channel; how much of that 5% revenue is likely to be lost if you close that channel; how real is that 5% number; what has the number been for different games; how constant has that 5% number been over time, for what period, with what trajectory; how constant is the 5% number likely to be in the future, or for other games that might be less popular or appeal to a different audience.
The key is whether you'd open up or lose a critical mass of customers by opening/shutting off a particular channel. Will customers comply with a push toward the higher margin channel? I'm sure Paradox lost some revenue by shutting down physical/retail distribution, but presumably the cost savings and higher margins of the online channel(s) justified that, (especially since the online presence probably expanded exposure and total revenue, anyway). Perhaps the costs truly justify this. But a less profitable channel does not equate to money losing. If Paradox could generate more revenue by restoring the physical/retail channel, and those would cover the additional costs, it would be in their interests to do so (again, unless it would enable/encourage many customers to migrate from higher margin channels to lower margin channels).
Frankly, though, the analogy to retail increases my skepticism. I know some retail distribution costs can be substantial. But I'm wondering how Paradox' QA and online distribution costs can be so high, and how the difference in margin can be sufficiently significant, that it'd make sense to risk even a small portion of that 5% revenue.