Company A would agree with you. But Company B disagrees, and so do I, personally. The biggest issue I can see with two guard switches in series is if one channel on one switch fails closed. If that guard switch is opened, then the safety relay will detect the fault, no problem. But what if the non-faulty guard is opened, and THEN the faulty one? The problem is not detected yet. Then, if the faulty guard is closed before the non-faulty one, the problem will stay undetected, and the safety circuit can be reset. This is clearly a problem! But that's where the usage of the machine comes in. If you have two gates, and the left gate closes over the top of the right gate, then two guards in series is a problem. Because you always have to open the left gate before the right, and you always have to close the right gate before the left. So if the right gate has a channel fail on, it can never be detected. Problem! But, lets say these two guard switches that are in series are not on a pair of overlapping gates, but on an infeed cover and an outfeed cover. Multiple times per hour, the infeed cover has to be opened to replace a roll of film/tape/etc, and multiple times per hour, the outfeed cover has to be opened to clean/adjust/clear something. In this case, you could, in my opinion, quite comfortably say that a fault on either guard switch will be detected very quickly. There's still a very slim chance it wouldn't be detected the very first time, and so I wouldn't consider this Cat 4 or PLe because your diagnostic coverage is reduced slightly, and Cat 4/PLe needs as high a DC as possible. But I'd consider Cat 3 to be reasonable, given it's wording around fault detection. You could also take into account how many operators run this machine. If there is only one operator, and the guards are at opposite ends of the machine, it's even less likely the fault would go undetected, as the operator will only be playing with one guard at a time. If there's an operator at both ends of the machine, that again changes things - it's now more likely that both guards will be opened at the same time, and by extension, slightly more likely that they will get opened in such a sequence as to mask the fault. Long story short, the way the machine is operated absolutely plays a part in the assessment.
Company B would agree with you, and personally, so do I. But that's the company line, and their employees have to toe it whether they like it or not. And Company A is one of the largest safety companies in the world, so it's hard to tell them they're doing it wrong. Having said that, they don't just ignore the design - as you say, if there are other means of detecting the faults they will accept it. We got around it in one case by using self-checking guard switches that do their own internal monitoring and detect faults internally, and Company A validated the machine with a dozen of those in series. But in general, they won't do it.
Again, all of this just goes to show that it there are tons of different interpretations and understandings of the legislation, and you can't hope for a clear cut, universal answer that will satisfy everyone.