Hi Emilie,
Thanks for sharing your library. I have taken a look and found a potential solution to the problem you are experiencing.
The way transition targets work is by modifying the probability of a transition for each primary/secondary stratum combination so that the expected amount of the transition matches your target. So for example if for stratum combination xy my target is 10 but the expectation is 50, all probabilities of the specified transition would be divided by 5 and the model would then apply the transition algorithm as it normally does. The process however is still stochastic. In your model, you are running spatially with dozens (hundreds?) of possible combinations. Add to that the fact that you are specifying a size distribution for your regeneration harvest transitions to mostly range up to 250 acres in size. So events may be starting in cells belonging to a stratum with a high target but spreading to a stratum with a lower target. Given all of the spatial constraints, it is difficult for the model not to exceed the targets because cells in strata with low targets have events spread stochastically into them from neighbors with high targets. I believe the problem is particularly pronounced in your model because you have so many small strata interspersed with each other such that these stratum crossing events become very common and all these small crossings add up to exceeding the target for some of your strata. Note that strata with a target of zero would still not have transitions spread into them from neighbors so multipliers would be redundant.
The easiest solution, is to not specify a size distribution for your regeneration harvest transition at all. In this case, the transitions will be simulated one cell at a time, without spreading to neighbors in other strata that may have very low probabilities. I have tried this solution with the example you sent and the model appears to track the specified targets closely.
Regards,
Leonardo