Garmin claims their units are verified against a "certified" reference device to be 5fps accurate at 2999fps and 2fps accurate at 300fps.
The communicated "precision" for the Garmins, as well as the LabRadar LX's and the Athlons, is supposed to be +/-0.1% for rifle cartridges (I'm not terribly certain why the precision specification would quadruple for slower projectiles, other than a potential for lower integrity frequency shift - but in theory, the units should be able to hit the projectile more times and do better interpolating velocities). "within 5fps on 2999" would be wider than +/-0.1% by almost double.
But... Given hundreds of rounds in testing with 6 or more chronographs operating side by side, and comparing the disparity between the two units of each brand, and between the units themselves, after ALL of those rounds fired and compared, Garmin is the only brand which could POSSIBLY be within their published specification for accuracy.
How I'm comfortable stating this - example: say I have a true velocity of 2807.0fps. One unit, landing within +/-0.1% of that, +/-2.8fps, could read 2809.8 as the absolutely highest reading which could still be within the specified accuracy, while another unit could read 2804.2fps, as the absolutely slowest reading which could still be within that specified accuracy. In that case, if I believe both are telling the truth, then I could know the 2807.0 is the true velocity, because no other velocity could be shared by the two while still achieving their published specifications. But more realistically, if the two units EVER displayed anything larger than 2x 2.8fps apart, larger than 5.6fps spread between the two units, then at least ONE unit MUST be reporting outside of the published specification for accuracy, in other words, a spread of larger than 5.6 between two units would suggest at least ONE unit is failing to achieve the specified accuracy to truth. In multiple 30, 50, and 100rnd experiments, the Garmins are the only units which have consistently reported velocities close enough together to have both units within the specified accuracy. The LabRadar LX's have been loosely just outside of their published accuracy on average with the max spread doubling the acceptable band, and the Athlons have been MUCH farther outside of their published specification for accuracy - with the Average difference between the two units often being farther apart than the specification would allow.
So this is a non-definitive method, such that ALL of them could be wrong, relative to true velocity, however, in my testing so far, with hundreds of rounds fired, the Garmins have been the only units which have the potential to actually be right. The Garmins MIGHT be reliably within their published spec for accuracy to truth, but by default, I can prove at least one, if not both, of the other brand units are NOT close enough together for them to both be within spec of truth.