When concept-based XAI is imprecise: Do people distinguish between generalisations and misrepresentations?