Exogenous hair dyes provide valuable information about the origin of a fiber since the applied colorants and products of dyeing process are able to preserve the relevant spectral fingerprint. SERS coupled with probabilistic methods allows the characterization of hair dye pathway and colorant composition, as well as estimating the degree of similarity between two dyed hair samples in terms of dye mixtures used with minimal hair sample consumption. DyeSPY-LINK converts similarity measures between hair dyed with the same dyes into likelihood ratios that represent evidential statements concerning common origin of dyes. A single LR, however, can convey misleadingly precise information when dye pathway classification, colorant prediction, and replicative sampling are considered deterministic operations. This paper develops an uncertainty-aware likelihood ratio framework that accounts for multiple error sources throughout the evidence evaluation process. Hair dye pathway and colorants become latent variables; spectra are resampled within strand-within-dye design; match and non-match score distributions are re-estimated in the Monte Carlo procedure; and LR distribution rather than a single LR value is provided as a result of the analysis. Validation in the new context is conducted at dye source level to avoid spectral leakage; measures of discrimination power, calibration, misleading evidence rate, uncertainty interval coverage, uncertainty partition, and decision category stability are proposed as quality indicators of the developed framework. Output LR is reported together with its calibrated log-scale uncertainty interval and posterior probability that evidential category remains unchanged after propagating uncertainty.