With a sufficient level of measurement resolution and precision, I think you'll find most pellets to exhibit some degree of ovaling. A lot of times it can be picked up with something as crude as a dial caliper (meaning approx. half-thousandth resolution), but sometimes needing a micrometer to discern.
As a singular variable looked at in isolation, I don't think heads being slightly out-of-round holds much predictive value as to how a tin of pellets will group. The spin imparted by the rifling is there to counteract it and other variables like the pellet’s weight bias (a little heavier on one side than the other) or slight geometric/aerodynamic asymmetry (a little ding in the skirt, some subtle mold flashing, etc.). Without the spin, these defects would cause each pellet to veer off in some apparent random direction. But by spinning it, the defect is constantly changing the direction it wants to veer off...just imagine it as an invisible hand that nudges the pellet at the 3 o’clock position, then at the 6 o’clock position, then at the 9 o’clock position, then at the 12 o’clock position. And around and around it goes, constantly being nudged back into some averaged middle trajectory.
Granted in severe cases it's surely an impediment but a deficiency in roundness calls into question overall quality control like the aforementioned geometric or gravimetric inconsistencies. After several years of measuring and weighing, I eventually realized the only way I've been able to improve my odds of getting an outstanding tin of pellets is to sample the better brands (JSB, H&N, RWS, Norma, QYS) at different times in order to get pellets made from different molds (or molds at different stages of their life cycle).