There is something really interesting about the fact you're grinding it up which I'm not sure has been mentioned above.
Imagine a batch of chicken. Perhaps one piece has Salmonella spp on it. For the supplier to find it, they will have to find that piece of chicken and test where the pathogen is on that piece.
Now think about your process. I'm imagining some bowl choppers, sorry for the metric but I'm imagining you're putting in, say, 40kg of chicken thighs into that bowl chopper. Then you're tipping it into a tote bin and putting it through something like a Vemag machine to extrude into skins. If it's not that process then it's probably going to be similar right?
I'm guessing you're not cleaning your bowl chopper until the end of the run and conservatively, let's make our run 5 batches (mostly to keep the maths easy).
So in total you now have 200kg of chicken thighs which have gone through the bowl chopper. And in that process, you're going to add in a bit of heat. Not deliberately but inevitably because of the friction of the process you have. So any pathogen presence has the chance to multiply.
So using nothing more scientific than a photo on the Walmart website (I'm wary of using British chickens to compare with this, as we tend to kill them when smaller) a chicken thigh seems to weigh about 160g or so.
In your 200kg of chicken thighs, you will then have about 1,250 individual thighs.
So your sample, especially if taken towards the end of the run is equivalent to a composite of over 1000 of their samples. No wonder you will find it more often. There is no way they are testing that much, am I right?
What's more is if you have one piece which is really grossly contaminated, the pathogens will then get dispersed to lots of lovely new food source without waste products etc. So what is likely to happen is an uplift in growth rate independent of the temperature.
Sorry that was quite long winded. But basically if you want to explain this, it's a combination of maths and the sampling method of the original material which can never be anywhere near the level of your product. You will always have a problem here in my opinion at least from time to time. To achieve the levels you want in your product, the supply is going to have to be super squeaky clean, I'd also encourage regular batch breaks, applying cooling on your process and regular break full cleans in your plant. (Also I'm not sure re US legislation on speciation but I'd never run from one species to another without cleaning.)
There are two other areas of investigation which I am going to share, but I don't think it's the cause in your case albeit just for your consideration to rule in or out. Machinery being the root cause of contamination can occur and can show up negatively in early batches. This is where there is water trapped deep in the machine which then leaches out very slowly during running. I've seen it happen with early batches passing and late in the day batches failing. As chicken in so synonymous with Salmonella, I still doubt it, but mentioning it for completion. If you have this kind of issue, using steam has been something I've seen and used to help at least "cook" residues deep in the machine short term but longer term you'd probably need some kind of repairs or replacement. It might be worth getting a full strip down PPM done on one of your machines you use to grind the meat just to see if there is debris, water etc getting deep within the machine.
My last suggestion, I am not sure if it's going to go down well with US companies. I've used either full genome sequencing or multi locus sequence typing before to help link with some pathogens. This may be worth looking at here. BUT my only two warnings are firstly, it's expensive and secondly it only shows you that the pathogens are linked, not how they're linked. So for example, if you find debris inside your bowl chopper or grinder which tests positive for Salmonellae and then you keep getting the same strain repeatedly, it could indicate it's your machine which is contaminated. Or it could indicate a harbourage problem in the evisceration plant. Or it could indicate endemic Salmonellae from one farm. But from US legislation having a WGS which links a certain Salmonellae strain to your plant I'd imagine would go down very badly with your leaders but the idea is out there. In this case though, even in the UK, I'd hold out and work with the supplier and shorten runs first.