Interpreting Log, Dummy and Interaction Coefficients

You fit the following model to 40,000 e-commerce sessions, where revenue is session revenue in dollars, pages is pages viewed, mobile is 1 for mobile and 0 for desktop, and returning is 1 for a returning customer:

log(revenue) = 2.10 + 0.08·pages − 0.35·mobile + 0.50·returning + 0.12·(mobile × returning)

1. Interpret the coefficient on pages as a percentage effect. Is the "multiply by 100" shortcut acceptable here?
2. What is the estimated percentage difference in revenue between a mobile and a desktop session (a) for new customers and (b) for returning customers, holding pages fixed?
3. A colleague reads the output and says "mobile reduces revenue by 35%, and returning customers spend 50% more." Give two distinct reasons that statement is wrong or incomplete.