Difference adjustment subtracts the roll gap from all prior history, keeping point moves correct but eventually producing negative prices on long histories of steeply contangoed markets like crude. Ratio adjustment multiplies instead, keeping percentage moves correct but distorting absolute point values.
This changes which statistics are trustworthy. On a difference-adjusted series, use point differences and not percentage returns. On a ratio-adjusted series, use returns and not absolute levels. Mixing them silently corrupts any volatility or z-score calculation.
Nothing is truly free here: no adjustment method gives you a series where both the level and the returns are simultaneously real. Keep the unadjusted contract prices alongside, and compute anything price-level dependent, like round-number stops, from those.
Related: contract-stitching, roll, contango, ohlcv-bar