I assume that the bevel gear is driven by another bevel gear. In order to calculate the angles, and other parameters, we need to know the number of teeth on both gears.
There are two methods of machining a bevel gear on a mill, with involute cutters, neither of which results in a perfect tooth shape.Straight tooth bevel gears are normally designed using an integer value of DP at the outer diameter.
The first method uses an involute cutter of the correct DP and tooth count for the outer diameter. But in order to pass through the gap at the smaller diameter the cutter is thinner than standard. These special cutters were often stamped 'BEVEL'. As far as I know these cutters are no longer available commercially. Note that normal spur gear involute cutters cannot be used. The method results in a bevel gear with a tapered tooth shape, but the wrong curvature (not enough) at the small end. This is normally corrected by filing.
The second method is the parallel tooth depth method. This is the method described in detail in the Ivan Law book. As it implies the tooth depth is constant rather than tapered as with a true bevel gear. These type of gears mate well with each other, but I suspect they will not mate properly with a conventional bevel gear. The critical design point with paralllel depth bevel gears is that the DP, and design calculations, are specified at the inner diameter. So a standard involute cutter can be used. However, from experience I can say that it is darn near impossible to design a parallel tooth bevel gear, with integer DP, to fit an existing mechanism intended for use with conventional bevel gears. At least not without major changes to the mechanism.
I suspect that JasonB's advice to re-use the gear if possible is the easiest solution. If I was faced with making the gears I'd CNC them, rather than use a conventional milling machine; been there, done that. I could probably make the gears, but it wouldn't be in the immediate future due to other pressures. PM me if interested.
A picture of the gear/mechanism would be very useful.
Andrew