The p-value of IQ differences was 0.21, which means it was insignificant. In other words, 21% of the time you would expect the same difference just by chance. A p-value should be below 0.05 to be clinically significant.
Randomization adjustments? If you're referring to the "adjustments for randomization stratum" (pages 1, 3, and 4), then you may not understand randomized trials. People are placed into groups of equal numbers, but some usually don't finish the study because they move away, they're non-compliant, etc. This results in a different number of subjects in each group. "Randomization adjustments" are mathematical ways of "equalizing" the statistical power of the groups as if they all contained the same number of subjects. If you look at Figure 1, there were 265 children measured for baseline IQ scores in the Amalgam
group, and 261 in the other group. If they both had 265 subjects, then this would be an unnecessary step.
You are correct that randomization adjustments were made, but this only strengthens the study by removing a potential skew.