arXiv:2611.23552v1 Announce Type: new Abstract: Reference Signal Received Power (RSRP) is widely used for beam selection in 5G NR as it is an easy indicator of path loss, but is not interference-aware, limiting its effectiveness in dense deployments. In this paper, we explore RSRP, Reference Signal Received Quality (RSRQ), and Signal to Interference-plus-Noise Ratio (SINR) as beam selection metrics alongside three UE-gNB association strategies in a 3GPP-compliant Urban Micro multi-cell simulation, benchmarked against Regularized Zero-Forcing (RZF) full-Channel State Information (CSI) upper bounds. Using a 3GPP-compliant Urban Micro multi-cell simulation benchmarked against RZF full-CSI upper bounds, we demonstrate two findings: directional RSRQ, despite its inter-cell interference awareness, is the worst-performing strategy in symmetric interference, and sequential association outperforms power-optimal Hungarian assignment. We also establish that the residual performance gap to full-CSI precoding stems from codebook quantization rather than metric imperfection, identifying adaptive codebook design as a mechanism to enhance performance.