The first thing to understand is that under normal graft conditions, the scion transformation will be about zero, with mostly just “physiological” type effects. Graft hybridization of the seedlings of a scion will likewise be low under normal graft conditions, however its inherently higher than the previous example, perhaps 0-15% novel characters depending on the species among other things. I would follow the technique for the mentor graft or you’re likely wasting time and effort.
It’s true there is still very much unknown about this. There is also a lot of misunderstanding in general about this subject. The research on graft hybridization is very interesting and there are different types of studies with different conditions and results. There are in vitro methods, but the ones I’ve read about all require inserting antibiotic resistance genes prior to making the cell grafts, in order to select out hybrids later, so a non starter for amateur breeders. I honestly wouldn’t worry too much about looking into the molecular mechanisms. The fact is now that there are an ensemble of different ones we know of that potentially are at play. With the level of complexity being so high, a full understanding is still beyond our understanding. Graft hybridization is indeed highly contingent, but it’s not random or unpractical. There are very powerful changes you can make to the phenotype/genotype of plants this way, and you can do it across different families (even grasses) if you follow the right protocol. You end up changing the regulation of thousands of genes among other things. Rather than thinking, “will the particular gene that I want get transferred into the seed” it’s more helpful to think “how can I blend to the degree that I want qualities from the rootstock into the scion and its progeny.” Here is some research done that you can get an idea. Its different from your typical piecemeal discrete thinking about genetic transfer.
Some examples:
Pear wide hybrid seedling mentor grafted on lemon rootstock becomes evergreen. This was a Michurin experiment. This is a 100% success rate int terms of transforming the scion in the direction of the rootstock. It may not be a 100% success rate in terms of making a particular discrete change that you want to make, but since you can effect it by degree and keep changing the direction of variation, then there are a wide open field of possible variations with some predictability. With woody plants of course you don’t need to look for variation in the seedlings since the graft lives many years in the field.
Tomatoes mentor grafted onto goji plants taking on the flavor and polysaccharide profile of goji fruits. Chinese research to make tomatoes more medicinal, this was successful and patents were subsequently filed. This is very similar to Michurin’s graft hybrids of pome fruits.
Vitis vines on shisandra rootstock showing a blending of various morphological characters from the rootstock increasing over the years. Also a 100% success rate in terms of every graft being transformed. The goal was to develop grapes with medicinal qualities of shisandra berries and was achieved. Again this is very similar to Michurin’s observations. This is a recent study with very detailed molecular analysis someone wants to look at it.
There are many more similar examples. There are also examples that more random and chaotic, such as novel phenotypic variation that wasn’t previously present. You find this in many studies, particularly the distant graft mutagenesis ones. But overall its a by degree type of phenomenon, which you can effect with by how long a plant lives grafted, or by how many repeated seedling generation mentor grafts you do, or both.
According to Michurin, the point of developing this method was two fold. Yes of course to introduce new variation not possible with sexual crossing, but moreover as a way of breeding for a small scale that is actually more efficient and less random than conventional fruit tree selection. The work is fairly painstaking, but it’s not the most technical thing either.