Hi Rob - the procedure you outlined is exactly how I do it, which is why my results puzzle me. For example, if you look at my plot for MGIV using under-the-lens Ilford MG filters (they were brand new at the time), my crossover is at a reflection density close to 0.3, so not far off from your data. What bothers me about that is, if grades 00-3.5 are speed-matched at ISO speed, the crossover should have been at a density around 0.6, not 0.3. With Classic, my results were further off, with a crossover point at a very low reflection density. This would actually be a more ideal speed-match point from my personal printing perspective than the ISO density, but I doubt my data is correct since in Ilford's tech sheet for Classic the crossover is clearly at a density of ~0.6, implying they are ISO speed-matched.
Perhaps it's simply a difference arising from the spectral output of the light source, but it just bothered me I was so far off on the crossover for Classic, since the curves themselves looked quite close to Ilford's.