This paper evaluates the performance of uplink integrated sensing and communication systems in the presence of gain and phase imperfections. Specifically, we consider multiple unmanned aerial vehicles (UAVs) transmitting data to a multiple-input-multiple-output base-station (BS) that is responsible for estimating the transmitted information in addition to localising the transmitting UAVs. The signal processing at the BS is divided into two consecutive stages: localisation and communication. A maximum likelihood (ML) algorithm is introduced for the localisation stage to jointly estimate the azimuth-elevation angles and Doppler frequency of the UAVs under gain-phase defects, which are then compared to the estimation of signal parameters via rotational invariance techniques (ESPRIT) and multiple signal classification (MUSIC). Furthermore, the Cramer-Rao lower bound (CRLB) is derived to evaluate the asymptotic performance and quantify the influence of the gain-phase imperfections which are modelled using Rician and von Mises distributions, respectively. Thereafter, in the communication stage, the location parameters estimated in the first stage are employed to estimate the communication channels which are fed into a maximum ratio combiner to preprocess the received communication signal. An accurate closed-form approximation of the achievable average sum data rate (SDR) for all UAVs is derived. The obtained results show that gain-phase imperfections have a significant influence on both localisation and communication, however, the proposed ML is less sensitive when compared to other algorithms. The derived analysis is concurred with simulations.