It's definitely true that Lenat won those two tournaments, and that the fleet design was computer-assisted. But optimizing a fleet would have been a relatively simple search problem, and he was the only competitor with a minicomputer. When the problem could have been solved by any number of simple algorithms, it's hard to take those solutions as evidence of a one-of-a-kind self-improving AI.
But Lenat isn't just saying that Eurisko searched the space effectivelyl. Instead there are all these stories about amazingly clever strategies that Eurisko found. But most of those stories appear to be incompatible with the actual rules of the game. And the stories aren't even self-consistent, but seem to change on every telling.
Here's some examples, mostly from tracking down the original Traveller High Guard and Trillion Credit Squadron supplements, and trying to map the stories to the actual game concepts:
Eurisko's design for ships that supposedly fired on themselves if damaged, to self-destruct and ensure that the minimum agility of the fleet did not drop? The rules only allow firing on the enemy. Could not have happened. Total fabrication.
The story about the "fuel tenders", that Eurisko made into warships? A minimum reserve was not actually a requirement in the rules. But if it were, there is just no chance that you'd need computer optimization would figure that out.
The story about the Eurisko's unhittable, unarmed, super-fast lifeboat that always guaranteed a draw? By the tournament rules a fleet that had no weapons left to fire was deemed to have lost. (In the link you posted with the fleet, I assume that this is the point of the Bee by abusing the Emergency Agility rule. But it should work out to a -8 DRM and thus hittable.)
The claim that he was told to quit the tournament or it'd be shut down? Disputed by the people who ran the tournament. Why would they need to threaten the cancelation of the tournament? They could just ban Lenat, or ban computer aid, with no repercussions. It's just pure embellishment.
The story about how all the opponents resigned after one round, one even before the first round, because it was so obvious that they'd lost? That's just absurd. Nobody goes to board game tournaments to resign without playing, and this is totally incompatible with the legend that Eurisko's designs looked strange and inefficient.
In terms of inconsistencies, when Lenat told the story in the '80s, his fleets were supposed to be slow and heavily armored. When he told the story to Malcom Gladwell 20 years later, he instead had an uncountable number of tiny, defenseless ships with big guns, and that the enemy could just not kill all of those tiny ships in time. A diametrically opposite story (and the latter one, as one should be used to by now, not possible by the rules since the rules capped the number of pilots for the fleet at 200, and even a small ship required at least one pilot).
Basically, when Lenat says something trivial but verifiable about Eurisko and these tournaments, it's highly likely to be obviously untrue. Just completely made up for a punchier just-so story. Why would we believe his unverifiable claims on non-trivialities, like the nature of his super-AI?
But I think the broader point is that even if it could self modify, Eurisko was incapable of learning. Like every other GOFAI program, it exists in its own carefully programmed universe of logic. And it's usefulness as a stepping stone to AGI is limited.
GOFAI research continues today under the label "cognitive architecture".