Between subgroups is not significant, but the replacement with PUFA is significant in itself.

As I’ve already pointed out to you, that’s not a contradiction when the studies are looking at CHO as a whole and not disambiguating between whole grain and refined carbohydrates.

If you think a 21% reduction in one of the largest killers in the western world is a tiny effect, then that’s genuinely laugh-out-loud funny.

Mechanistically, replacing saturated fats with PUFA reduces the amount of arachidonic acid in cell membranes which dampens excessive inflammatory cell signalling. Here's how that works. "Because arachidonic acid (AA) competes with EPA and DHA as well as with LA, ALA and oleic acid for incorporation in membrane lipids at the same positions, all these fatty acids are important for controlling the AA concentration in membrane lipids, which in turn determines how much AA can be liberated and become available for prostaglandin biosynthesis following phospholipase activation. Thus, the best strategy for dampening prostanoid overproduction in disease situations would be to reduce the intake of AA, or reduce the intake of AA at the same time as the total intake of competing fatty acids (including oleic acid) is enhanced, rather than enhancing intakes of EPA and DHA only. Enhancement of membrane concentrations of EPA and DHA will not be as efficient as a similar decrease in the AA concentration for avoiding prostanoid overproduction." https://pmc.ncbi.nlm.nih.gov/articles/PMC2875212/ The same authors also noted that "The degree of fatty acid unsaturation of mitochondrial membrane lipids has been found to be one of those biochemical parameters that are most strongly correlated with longevity, when different species of mammals and birds are compared, with a low degree of fatty unsaturation being correlated with less lipid peroxidation and a longer normal life-span." So, mechanistically speaking, replacing saturated fatty acids with PUFAs will dampen inflammatory cell signalling. However, replacing saturated fats with monounsaturated fatty acids will do the same thing without risk of peroxidative damage to cells. Excerpt: "Fatty acid composition in the Western diet has shifted from saturated to polyunsaturated fatty acids (PUFAs), and specifically to linoleic acid (LA, 18:2), which has gradually increased in the diet over the past 50 y to become the most abundant dietary fatty acid in human adipose tissue. PUFA-derived oxylipins regulate a variety of biological functions. The cytochrome P450 (CYP450)–formed epoxy fatty acid metabolites of LA (EpOMEs) are hydrolyzed by the soluble epoxide hydrolase enzyme (sEH) to dihydroxyoctadecenoic acids (DiHOMEs). DiHOMEs are considered cardioprotective at low concentrations but at higher levels have been implicated as vascular permeability and cytotoxic agents and are associated with acute respiratory distress syndrome in severe COVID-19 patients." https://www.pnas.org/doi/10.1073/pnas.2120691119 More info in COVID-19 risk: "Separately, on analyzing global COVID-19 mortality data and comparing it with 12 risk factors for mortality, they found unsaturated fat intake to be associated with increased mortality. This was based on the dietary fat patterns of 61 countries in the United Nations' Food and Agricultural Organization database. Surprisingly, they found saturated fats to be protective." https://www.medpagetoday.com/reading-room/aga/lower-gi/86940

Why would we care about mechanistic speculation from in vitro models when we have evidence from higher up the evidence hierarchy showing direct benefits on human health outcomes when PUFA is substituted for SFA?

It’s like someone providing evidence that exercises decreases mortality risk in humans and countering with “but exercise induces oxidative stress at a cellular level so we shouldn’t do it.”

It’s just turning the evidence hierarchy on its head. I’m curious why you would interpret evidence in this way - why are mechanistic studies with incredibly low internal validity more convincing than RCTs and prospective cohort studies to you?