The vulnerable groups resulting from transition recession in low-income countries were not always able to rely on state-subsidized services or cash safety nets. Informal transfers from clan and community members became in many cases the dominant coping mechanisms for the poor. In these country contexts, a mixed targeting of means testing and community-based has been considered a good strategy for the effectiveness of social programs. The only anti-poverty programs in Albania (the Ndihma Ekonomike) used this strategy with a resulting marginal impact on poverty and persistent undercoverage and leakage errors. Previous analyses on this program suggests good program's ability in reaching the effectively poor. This study performs an impact evaluation of Ndihma Ekonomike finding a negative impact of program participation on welfare. As evaluation method, we apply a regression-adjusted local linear matching in order to minimize bias due to program's selection. The model considers also the targeting errors in the following ways: controlling for targeting performance at local level and performing the analysis also on sub-samples of well-targeted households.