Just remember to avoid the misconception there is always a single critical path, real life contracts usually involve several delivery milestones.

From

http://www.icoste.org/AACE2008%20Papers/Toronto_ps07.pdf

The longest path basis for determining the critical-path is flawed, because of the common use of internal start-no-earlier-than and finish-no-later-than date constraints. From a practical standpoint, the prolific use of date constraints cannot be mandated out of existence, and is likely to only increase in the future. Actually, we encourage the discrete use of date constraints as a practical and effective way to simulate project management intentions and their likely consequences.

Although I completely agree with the above statement I would like to add that some software wrongly implement the FNL constraints by fixing late dates to impossible dates, ridiculous late dates early than early dates !!! thus creating the negative lag aberration. Not all software implement the modeling of FNL constraints so wrong with such distortion of the float values, one of the most important metrics in scheduling.

It is simple, if a FNL constraint cannot be met then the late dates become equal to the early dates, late cannot be earlier than early, a missing mathematical constraint in such software. Note that the right mathematical model Keeps the correct identification of criticality on these activities maintaining float equal to 0 and not by assigning dates that have no meaning in the physical world as we know it.