I figure you can't really adjust Roman coins or something of that ilk for inflation to the modern era, so there must be some point where adjusting for inflation gets tricky and money gets weird because currency changes.
Even in the modern day, it can be quite difficult to compare inflation statistics between countries. Inflation is calculated using a Consumer Price Index; which measures the price of a "basket" of goods and services that a "typical" citizen consumes over the course of a year. However, what gets included in this "basket" depends on who's preparing the statistics.
In the US, the CPI has been compiled by the Bureau of Labor Statistics since 1913. The "basket" of goods is periodically updated periodically to reflect changes in American spending habits, and the methodology of how the CPI is calculated has also changed throughout the years. Other countries use their own methods to calculate their CPIs, and various international bodies and publications will each calculate CPI differently when comparing inflation rates between countries.
When writing about historical prices, historians have used various indirect comparisons; such as comparing the typical wages from the period in question with modern wages or comparing the prices of basic goods (food, clothing, etc.) that are still commonly used today. Although these methods can give you a rough idea of how expensive something was, you can't use them to adjust for inflation (in the modern sense of the term), there simply isn't enough data.
Edit: In the US at least, you can adjust for inflation using modern methods back to 1913. Even though the exact methodology have varied over the years, the BLS statistics are considered a reasonably reliable gauge of inflation in the US.
Would inflationary measures with their 'basket' take into account that many products (for recent example, HD tvs and smartphones) simply didn't have precedents?