Razor's exactly right. Statistically, you can't address probability concerns (as you've done) without studying the problem over time. YES TIME, always LOTS of time.
It's essential for you to perform a daily tally of the reward achievements each day for your buildings, over a reasonable duration of time. Reasonable isn't a few days. The longer you test, the more the tally of choices (and derived %) should align with what's stated in the building. It's tough to invest 3 months of daily scorecarding. But you'll gain credibility for your problem if you can prove your point by sound math. That's the investment you need to make.
IF not, we can't get behind your hypothesis without you demonstrating reasonable investment in discovery. If you embrace the intelligent suggestions from multiple people about the "math of this"... you learned something about probability over time.