We focus on the impact of failing to control for differences in land types defined along toposequence on estimates of farm technical efficiency for small-scale rice farms in eastern India. In contrast with the existing literature, we find that those farms may be considerably more technically efficient than they appear from more aggregated analysis without such control. Farms planted with modern rice varieties are technically efficient. Furthermore, farms planted with traditional rice varieties operate close to the production frontier on less productive lands (upland and mid-upland), but significant technical inefficiency exists on more productive lands (medium land and lowland).