This paper proposes a flexible and enhanced hybrid three-stage model for efficiency measure，ＲAMSFA-RAM，without the mandatory adjustment of the effects of environment factors and statistical noise，which is based on the comprehensive discussions and review of extant hybrid three-stage models for efficiency measure with adjustments for the effects of contextual environment and statistical noise． In contrast to the existing models，our hybrid model not only obeys the non-linear productive functional relationship between the environmental factors and slacks in estimating slacks and in adjusting the inputs and outputs to reduce the bias of efficiency estimation，but also takes advantage of the ＲAM which can present a new approach of adjusting the different effects of environmental factors and statistic noise based on both inputs and outputs． More importantly，our model takes full advantage of the translation invariance of the ＲAM to deal with the non-positive values，which can only work for existing models by arbitrary positive adjustments in virtue of the maximum or minimum of estimated slack datasets，and avoids the bias of efficiency estimation． In the empirical study，this article applies this new model to measure the research and development efficiency of large-and mediumsized industrial enterprises at China’s provincial level，to show that our method presents a feasible modelling approach of objectively comparing efficiency scores．