We consider the Grover search algorithm implementation for a quantum register of size N = 2 k using k (or k + 1) microwave-and laser-driven Rydberg-blockaded atoms, following the proposal by Mølmer, Isenhower, and Saffman [J. Phys. B 44, 184016 (2011)]. We suggest some simplifications for the microwave and laser couplings, and analyze the performance of the algorithm for up to k = 4 multilevel atoms under realistic experimental conditions using quantum stochastic (Monte-Carlo) wavefunction simulations.