* NLA scaling was being incorrectly handled with fractional values. This was caused by clamping to integers instead of floats when making sure the scale value was positive.
* Added checks to help prevent some weird cases that may sometimes occur and cause problems