I'm sure this has to do with the differing attack/release characteristics of the hardware vs the software...
I thought just the same. My first impression was that the hardware units had a faster response in attack/relase or at least a less noticecable curve, I mean, I´m not sure if it´s just a question of atack/release time or if it´s more like if were a more progresive curve that makes less noticecable when starts to compress.
On the Sofware samples, specially the Acoustic one, it´s very evident when the compressor enters. And seems go sudenly from no compress to a 2:1.
In the hardware seems like if the threshold point were more large, and so if like start the compression at , 1.2:1 1.5:1 or something like that.
I have another theory, maybe the hardware compress separately the highs and the low frequencies? something like an internal multiband compressor? All this a all speculations I´m thinking now since I don´t know how works the La2A hardware unit. You´ll better know that.
I thought about this because usually, sofware compressors, even the best ones, the high frequencies often suffer more, or became less life, and dull.
I think any type of multiband must be needed, because sometimes we end compressing high frequencies that are already on low level. And the threshold the first it sees are the lowest frequencies, or if you want , the fundamental wave, what are usually at much higher level that the harmonics.