The key to split grade printing lies in the contrast of the negative. A very thin negative will not produce a good print using split grade printing because it generally lacks contrast and therefore requires a harder grade to build the contrast required in the final print. When I am faced with such a negative on a workshop or for a client I inevitably make the print using only one grade, either 4 or 5. A so called normal negative will provide a better starting point because there should be contrast and detail throughout. However, my own view is that a negative with higher than normal contrast is the ideal starting point. I reached this view after spending many hours trying different combinations of negative contrast and using different combinations of high and low contrast filters. I finally decided on a higher than normal contrast for the negative, grade 0 and grade 5 filters only and accurate control of time given when exposing with each filter. I use an RH Designs timer and work in fstop mode frequently working to 2/10ths of one second, and it does make subtle but significant differences in the final print.
The reason that the high contrast negative works better than normal contrast is that the negative acts as a mask when exposing the grade 5 or hard filtration. Think about it, when you print with a grade 5 the highlights are frequently paper base white and part of the reason is that the density of the negative has prevented light from passing through the negative. Clearly, this is not the only reason for paper base white highlights but bear with me and I'll explain how to make this work for you.
My method of split grade printing is based on making 2 test strips and there is no guesswork involved, you respond to what you see. The weakness in some of the methods used by many printers, such as the matrix method where you expose your test strip one way with soft and the other way with hard filtration is that only one small square somewhere on the paper is correct. equally, starting with 2/3rd of the exposure on hard and 1/3rd hard requires some guesswork or at best more test strips to arrive at the desired result.
I make a test strip using grade 0 and select the exposure that gives me the tonality I require in the highlight. At this stage I totally ignore the shadows and overall contrast, there is none because you are using a soft grade. Having chosen the exposure that gives the highlight tonality, I expose the whole of the second test strip at that time and change filtration to grade 5, and expose a series of increments over the grade 0 exposure. When processed the second test strip will show good contrast with detail through from highlight to shadow. You will see subtle changes in contrast across the test where you have given more exposure with grade 5. Doing the test this way gives you accurate information from which you can make the choices required to produce the contrast you wish in the print.
Ann makes the point in her post that it works for her with negatives of extreme contrast and also comments that it works regardless of which filtration is used first. That is correct but there are subtle differences if the hard filtration is used first rather than the soft, the final print will be slightly higher in contrast.
I have two golden rules when split grading, they are: be careful not to give too much soft exposure. In 10 years of teaching my method the most common problem is that there is a tendency to give enough soft exposure to try to establish the lower values and this results in very muddy prints. The second rule is when making the print expose the two different filtrations in the same order as when you made the test strip.
I do not claim that this method id the only way to split grade print but it has worked for me and when I have taught it and written about it I have had a very positive response that it has generally helped photographs produce just what they are after.
Try it with one of those very high contrast negatives that we all have and have given up on. Sorry to be so long winded but split grade printing is not a simple technique to describe briefly. I've writted about 50,000 words on the subject over the past few years and still don't think that I've fully covered every nuance.