Unveiling Cultural Blind Spots: Analyzing the Limitations of mLLMs in Procedural Text Comprehension