The problem with using only certain parts of the body to get the scale is that the model companies don't always or hardly ever get the proportions correct.

If one is interested in working out a model's scale, it's worthwhile checking how the figure's proportions compare to those of the real animal. For this new Ankylosaurus toy, the relative size of the head, body and tail look like they will correspond well to the reconstruction in the new paper. The toy also appears to have very similar proportions to this Ankylosaurus skeletal:

https://getawaytrike.deviantart.com/art/Heavy-armored-special-615655175That Ankylosaurus skeletal's size is for the largest Ankylosaurus specimen, and by my calculations that skeletal shows it as being 7.04m long. Although I measured it in a straight line from the front of the snout to the end of the tail, and I think dinosaur length estimates tend to end up being somewhat longer due to being measured as if the bones were laid out in a row on a flat surface meaning curves in the body shape aren't taken into account. The toy seems to have a slightly longer tail than the skeletal, so that could maybe extend the length to around 7.5m. Based on the proportions of the toy and the reconstruction in the new paper, I don't see Ankylosaurus being more than 8m long. This is suggested by the paper too: "a body length of nearly 10 m for a large Ankylosaurus is probably too long, but a length of up to 8 m is probably within reason." Do you see how those longer estimates, especially 9 - 10 m would require parts of the animal such as the body or tail to be considerably longer than they are in certain reconstructions such as that skeletal?

So let's say the head is in a 1: 40 scale but the rest of the body is way to small. So you get another figure in that same scale to go with it but they end up looking completely wrong together but at least their heads will be to scale.

I think this is what's more likely to happen: A length estimate will be used to calculate a model's scale, but the length estimate is dependent on unknown lengths of body parts e.g. the tail. The result is the model will look overall too big/small compared to other models in that scale due to having different proportions to those used for the length estimate.

As an example of this, you calculated the scale of the new Ankylosaurus toy as being 1:45 scale if based on the new maximum size estimate. However, I've calculated it's around 1:36 scale for a 7.5m Ankylosaurus, or 1:38 scale for an 8m Ankylosaurus. If the toy's proportions are indeed similar to the new reconstruction in the paper as well as the skeletal, as I mentioned previously, I'd say longer length estimates for Ankylosaurus would not be correct for this toy and using them to calculate scale would just result in the toy appearing to be a smaller representation than it actually is.

Going by the length isn't perfect either because it does not take into account how fat or tall it is and if the tail and neck are curved that could really mess up your calculations too. But length is the simplest and easiest way to get a general idea of how your figures scale up with each other. But if your concerned about getting the exact scale I would look at both.

I think calculating scale based on overall length can work well for species where the overall length is well understood. For species where the overall length is ambiguous, I think comparison to known parts of the animal is important, but it can be of interest to also compare the model to an overall length estimate, and then see how the scale based on length compares to the scale based on known parts.