Typically, a Kalman filter is only as accurate as your mathematical model of the underlying phenomena (for example, your flight dynamics on an airplane), and the filter is mainly useful to mitigate the noise from your real-world sensor observations. Is there more public information available about the mathematical model being used within their state transition matrix? Depending on that information, this could be really clever or a glorified low-pass filter.