One problem with modern portfolio theory (MPT) is how easily the suggestions vary depending on the start, end, and asset classes included. Running a calculation from 1970-2010 will probably be different than the past 25 years (1993-2017), especially if you split stock allocation into: US large, US small, international developed stock, and emerging markets.
I think there's plenty of evidence that U.S. and international stock markets have become more correlated over time (as international businesses become more common and intertwined). With MPT, everything relies on the correlations. When correlations change over time, allocations change along with them. Which means data from long ago, when U.S. and international were less correlated, may not be relevant today. And could also explain why MPT provides varying answers based on start and end dates (when you have several stock asset classes, like emerging markets and US small cap).