Both of the calculations are just approximate I would say. I don't know how is either of them calculated in detail, co it's hard to tell if one of those is better.
But, have in mind, that usualy both have several parameters like salt and oligo concentration, that you can set. Maybe they use different settings or maybe they don't use some of the concentrations at all. It would be logical to expect that different enviroment of oligos have different effects on their self-alignment.
Also some of the polymerase mixes may have more problems with dimers than others (depending on the enviroment thing again, but it means that using some polymerase, primers will work better than in other). For example hotstart PCR prevents unspecific amplification in general, ale dimers are unspecific in their nature (the complementarity is not 100%). The higher specificty, the less dimers.
But, I suppose in any case, primers with high self-binding will "work" most of the time. The self-binding tells you how much they like themselves instead of the template. In different conditions again, the dimers will form more. For example, I thing using too high primer concentration:template ratio, will lead to preferential formation of dimers (since there is overabundance of primers), and dimers will form more likely, and when they do, they use up the polymerase and nucleotides, so less of it is available.
But to sum it up, all this thought excercise, when you design in IDT and it tells you it's fine (but in my opinion, IDT images of dimers is usually pretty confusing, you must know what deltaG is fine and which not, using primer3 (which is the algorithm in PrimerBLAST) is simplier, that it only tells you number, you know that high is bad, and try to avoid it) it's probably going to work. If you design primers in primer3 and use their recommended setting, it's probably also going to work.
Regardless on what different algorithms say.
And just to make it more complicated. I personaly seen primers, that have so bad in silico parameters that I though they never gonna work. And they did. Because all these tools will never tell you which primers will work, but they will try their best to point out which has higher probability to not work. And they still may, specific sequence, specific conditions.. these ale all unpredictible.
So try to design "good" primers, but it's not a rule that "bad" will never work (or that "good" ones will). Just probability.