I read MJK's paper on OB speakers and I am going to embark on a Fostex, Eminence build. Are the crossovers Martin describes 2nd order Butterworth? He has Caps at 24 and 68uF and coils at 4 and 9 mH. I can only match these to Butterworth. Should the XOs be LR?
Not only at the XO frequency, a larger picture should be taken into acount. The XO circuits and drivers are interacting full band. An ideal method is to flatten the impedance of the drivers first, then do the XO jobs. You'll need Zobels for the rising towards the high, LCR notch for the low end resonance... etc. Sometimes more (unfortunately).
In the end, electrical curve shape doesn't matter at all. What you should aim for is the acoustic results. It's just too complex to describe in one or two posts. Issues in drivers themselves, baffles (or boxes), the interactions between them... all over the place.
Doing XO design from scratch and by pen/paper/calculator/ears, you'll need a LOT of time. I'd suggest getting a tool to measure the impedance of the driver, and a mic+RTA to see what really happens in acoustics. Those are cheap and easy to get now. You'll be easier and much more successful with them.
Martin only notes the xover frequencies and slopes in his paper and not what type they are. However, in the MathCad worksheet that he distributed as an example of an OB with an FE103E and Alpha 15A uses a LR filter for a low pass and Bessel for the high pass -- he did not go into to any detail about the xovers in his paper.
Martin only implements "textbook" crossovers in his worksheet and you can simulate the use of a Zobel by setting the impedence = 0. However as, CLS notes, you need to enter a considerable amount more of driver specific behaviour before being able to accurately predict how they might behave. Martin's worksheets aren't intended to be used for that.