A probability system or computer system setup to provide 'imitated' randomness would do so by ensuring that certain controls were in place to ensure the correct statistical probabilities were met over a certain range of integers.
So from an observational viewpoint, when that range of integers (or in your case shuffle) were observed and cataloged it would give the illusion that the system is in fact random.
However, here is the snag. The system itself, and observational viewpoint only focuses, or directs focus onto the range of integers with which the statistical adherence lies. Any deviance from the percentages on an individual basis (ie your individual shuffle) would not be highlighted. Therefore the system could manipulate the individual numbers, providing it is compensated for in a different entry.
So, if all the information you have does show statistical adherence, it only proves that over a fixed number of hands their systems are working to the percentages that are given to it. Not that it is in fact random, nor does it eliminate manipulation.
The final outcome of the hand is what actually counts towards the adherence percentage, and is what all monitoring and analysis of their systems will rely on. This is where the problem lies; the reporting of data is characterised not by the construct of a hand, but the final outcome. So any manipulation will not be reported, and cannot tracked by any third party systemic observations.
In order to highlight any manipulation that may, or may not be taking place, we need to investigate the hand information to show statistical data for each hand involved at every stage of the deal. A complicated and lengthy process as i'm sure you understand.