The Elektor publication, "Build your own High-End Audio Equipment" is a compendium of earlier Elektor designs published in their monthly magazine. Hence, in India, the Phase Linear Crossover article appeared in October 1986 and has also appeared in the former book.
The design uses fourth order Linkwitz-Riley filters (Low Pass) and also All Pass to simulate the time delay and then applies subtraction to obtain Band Pass and High Pass. This technique was an adaptation of the same used by Lipschitz and Vanderkooy.
Elektor also published another Phase Linear Crossover using only the subtraction method, but this time using fourth order L-R High Pass filters. If memory serves me right, this appeared in January 1985 in the Indian publication. This has however, not been included in their compendium.
The point is that I have tested both filters and have working boards even today. The filter that I have later referred to, has better sonics that the earlier mentioned one. But there seems to be a 'hole' somewhere about the crossover point. The one that Promitheus is referring to, which I have mentioned first, suffers from a slightly 'smudged' sound. I guess this may have to do with group delay. The dynamics and slam are present but the mid-band is slightly smeared. This is discernable only when compared with a 'better' filter. On its own, the Elektor design in worth the try.
The 'better' filter that I have mentioned is an adaptation of a circuit featured in Electronics World and was titled "Precise X-over". This was a two way design, that I configured as a three way. The simulations I performed on Electronics Workbench were exactly the same as published in terms of slope and phase. The slopes are between fourth and eighth order, something like sixth order. Within the pass band, the phase remains the same and group delay is minimal, since second order L-R filter configuration is used. The sonics of this filter is out of the world. I use them with a full horn loaded three way system for outdoor use and the results are more than just impressive. And to boot, I am only using a combination of LF353 and NE5532 opamps. Better opamps would probably sound even better.
Give this a try, seriously.