Recent years have witnessed the development of the Metaphor Identification Procedure (MIP/VU), a step-by-step protocol designed to identify metaphorically-used words in discourse. However, MIP(VU)'s merits notwithstanding, the procedure poses a problem to scholars intending to use its output as the basis for a semantic field analysis involving a quantitative component. Depending on the research question, metaphor analysts may be interested in chunks of language situated above the procedure's standardized level of analysis (i.e., the lexical unit or lexeme), including phrases and sentences. Yet, attempts to decenter the method's exclusive focus on metaphor-related words have been the target of critique, among others on the grounds of their lack of clear unit-formation guidelines and, hence, their inconsistent unit of analysis and measurement. Drawing on data derived from a Spanish-language US-based newspaper's coverage of the migration program known as DACA (Deferred Action for Childhood Arrivals), this article describes challenges that analysts can run into when attempting to use a dataset containing atomized metaphor-related words as the input for subsequent quantitative semantic analyses. Its main methodological contribution consists in a proposal and illustration of three possible methods to extend the existing MIP(VU)-protocol in such a way as to allow it to capture metaphorical strings, on top of lexemes, in a reliable and systematic manner. The first two methods are procedural, and entail formulating a-priori grouping-directives based on the research question(s). One departs from semasiological criteria (Method 1) and the other takes an onomasiological approach (Method 2). The third method works bottom-up, involving the ad hoc grouping of lexemes and adding a descriptive parameter meant to keep track of grouping-decisions made by the analyst, thereby safeguarding transparency at all times.