The thickness of paint films is one of the most critical quality parameters in the aeronautics paint process. Numerous measurement techniques exist for quantifying coated film thickness but many are unsuited to deployment in industrial environments or non metallis subtrate. With the advent of robust, turnkey systems in recent years, time-resolved terahertz pulsed sensing has matured sufficiently to find application in providing quantitative analysis of physical properties to a range of industries, including semiconductor package inspection [1] and non-destructive testing of larger scale composites For many applications, layer thickness determination can be accomplished using a simple time-of-flight approach, whereby the thickness of individual layers can be estimated from the successive reflections from front and back interfaces of each layer[3]. This may be applicable to thick paint layers where the interface reflections are well separated, as in marine paint systems [4], but modern paint systems such as aeronautics painting are comprised of many thin paint layers, making accurate identification of individual interface reflections difficult[5]. Terahertz time-domain spectroscopy is a specifically appropriated technique to analyze layered structure composition and geometry and dimension. Inverse electromagnetic problems are commonly performed to extract, from a recorded reflection THz-signal, the distinctive layer dielectric properties and thicknesses. However, main origins and formation routes of the signal are left unassessed while it could be of great value to deepen, to enlarge and to optimize stratified material THz-probing possibilities, control and understanding. In this work, an iterative algorithm implements a connected propagation tree where each node denotes a subsequent pulse subdivision. For each interface, nodes are pre-filled with the respective dielectric interface transmitted and reflected power coefficients. Descendant pulses are thus individually monitored and their carried proportion of the incident power remains at any time accessible. By constraining the algorithm to a given number of internal subdivisions, the original signal is recovered and the power contribution of each optical path is assessable. On the basis of this knowledge, the transfer function to be calculated within inverse electromagnetic problems can be simplified to enhance algorithm convergence speed. For instance, convergence speed-up would benefit the automation thickness chain control in industry. The algorithm robustness is demonstrated through the reconstruction of a THz-signal from a micro-metric layered structure used in aeronautics. The power proportions of different optical paths contributing to each part of the signal were explored. The algorithm allowed to reduce by about 99\% the number of calculated pulses compared to a standard stratified model. Acknowledgement: The authors thank Nouvelle Région d'Aquitaine and BPI France for Athermo project funding