I changed the non_worker_multiplier from a uniform distribution from -0.05 to 0.15 to a much more optimistic uniform distribution of 0.5 to 1.0. The predicted outcomes still greatly favor Basic Job to the tune of ~1.5T.
I think that this demonstrates to a large degree the author's point that the best way to disagree is to make a change and see what you come up with. For someone with inclinations similar to yours, your comment is likely enough for them to justify dismissing the result reached by the basic model, but when you add the proposed effect to the model, well, you come to the same basic conclusion, that Basic Income is significantly more expensive than Basic Job.
A lot of comments here, including yours, come from an unjustified implicit rejection of the main thesis of the article by preferring rhetoric over model refinement. That is, the author is arguing that a better way to go about this discourse is to test the effects of different assumptions with the model, but many commenters are hammering through rhetorical rebuttals to the model's conclusion without addressing whether they actually believe that rhetoric is more effective than the mode of discourse proposed by the author.
Your rhetorical argument (including the argument of the grandparent since you apparently concurred: "absolutely") suggests that the main factor in deciding between the two approaches is whether they decrease or increase work productivity. When that hypothesis is tested in the model, it turns out that the other factors combined overwhelm the effect on worker productivity, and so the effect on worker productivity should not dominate the debate as suggested by the grandparent.
This was an easily testable hypothesis given the model offered, and once it was tested it looked weaker than suggested by the grandparent. I think that this demonstrates that there is some truth to the author's claim that running our assumptions through the model, and refining it overtime is likely more valuable than making these untested rhetorical arguments.
Now, certainly there are likely problems elsewhere in the model, and reasonable people can have disagreements on what assumptions are reasonable. But the point is to argue and test those assumptions continually, rather than offering unweighted rhetorical arguments and either never reaching consensus, or just landing on the argument that is most repeated or shouted the most loudly.