This paper addresses the issue of bias in Graph Neural Networks (GNNs) by introducing Real Fair Counterfactual Graph Neural Networks+ (RFCGNN+). GNNs have become crucial in various decision-making scenarios due to their superior performance, but they can inherit and exacerbate biases, particularly against marginalized groups. Most existing fair GNN approaches focus on statistical fairness, which often fails to account for labeling bias, where societal biases influence data collection, leading to distorted training data. RFCGNN+ introduces a novel model that identifies authentic counterfactual samples within complex graph structures and mitigates labeling bias through causal analysis. It also incorporates a fairness-aware message-passing framework with multi-frequency aggregation to address topology bias. Extensive experiments on real-world and synthetic datasets demonstrate the effectiveness and practicality of the proposed approach.This paper addresses the issue of bias in Graph Neural Networks (GNNs) by introducing Real Fair Counterfactual Graph Neural Networks+ (RFCGNN+). GNNs have become crucial in various decision-making scenarios due to their superior performance, but they can inherit and exacerbate biases, particularly against marginalized groups. Most existing fair GNN approaches focus on statistical fairness, which often fails to account for labeling bias, where societal biases influence data collection, leading to distorted training data. RFCGNN+ introduces a novel model that identifies authentic counterfactual samples within complex graph structures and mitigates labeling bias through causal analysis. It also incorporates a fairness-aware message-passing framework with multi-frequency aggregation to address topology bias. Extensive experiments on real-world and synthetic datasets demonstrate the effectiveness and practicality of the proposed approach.