If you use the t-statistic instead of the z-statistic, n = 250 => df = 249 which is effectively df = infinite as far as the values used in the calculation are concerned.

You'd use 0.99 because the tables give you Pr(z < z*), not Pr(z > z*). So you want the value z* such that Pr(z < z*) = 0.99, which means going backwards from 0.99 in your tables.

I'm not sure where the0.49comes from ......