Text this: Analyzing Bias in LLM-Augmented Knowledge Graph Systems