LiF:Mg,Ti, are widely used to estimate absorbed-dose received by patients during diagnostic or medical treatment. Conveniently, measurements are usually made in plastic phantoms. However, experimental conditions vary from one group to another and consequently, a lack of consensus data exists for the energy dependence of thermoluminescent (TL) response. This work investigated the energy dependence of TLD-100 TL-response and the effect of irradiating the dosimeters in different phantom materials for a broad range of energy photons in an attempt to understand the parameters that affect the discrepancies reported by various research groups. TLD-100s were exposed to 20-300 kV narrow x-ray spectra, (137)Cs and (60)Co photons. Measurements were performed in air, PMMA, wt1, polystyrene and TLDS as surrounding material. Total air-kerma values delivered were between 50 and 150 mGy for x-rays and 50 mGy for (137)Cs and (60)Co beams; each dosimeter was irradiated individually. Relative response, R, defined as the TL-response per air-kerma and relative efficiency, RE, described as the TL-response per absorbed-dose (obtained through Monte Carlo (MC) and analytically) were used to describe the TL-response. Both R and RE are normalized to the responses in a (60)Co beam. The results indicate that the use of different phantom materials affects the TL-response and this response varies with energy and material type. MC simulations reproduced qualitatively the experimental data: a) R increases, reaches a maximum at ~25 keV and decreases; b) RE decreases, down to a minimum at ~60 keV, increases to a maximum at ~150 keV and after decreases. Independent of the phantom materials, RE strongly depends on how the absorbed dose is evaluated and the discrepancies between RE evaluated analytically and by MC simulation are around 4% and 18%, dependent on the photon energy. The comparison between our results and that reported in the literature suggests that the discrepancy observed between different research groups appears to be most likely related to supralinearity effect, phantom materials, difference on the energy-spectra and geometry conditions during each experiment rather than parameters such as heating-rate or annealing procedure, which was supported by MC simulation. From the results obtained in this work and the strict analysis performed, we can conclude that for clinical applications of TLD-100, special attention must be taken when published data are used to convert TL calibration curve from (60)Co to low-energy photons. Otherwise, this can lead to incorrect results when later used to measure absorbed dose in human tissue.