I think matched filters are mainly used for ease of analysis; they make more sense from a filtering/convolutional perspective. Because both yield the same result and both have the same computational complexity, there is not much in implementation/performance that changes between the two.
Also, iVenky, I do not think the two are different. The integrator in the correlator is from 0 to T, so a proper correlator will only consider that window, which is the same as the matched filter window.