Brian provided a file that showed where the DEPEND_0 and DEPEND_1 datasets didn't pick up there ISTP metadata when VAR_TYPE was not included.
This is because the variable which depended on these was treated as ISTP because it contained a DEPEND_0 property, but the two DEPEND_0 and DEPEND_1 datasets did not. This is fixed by allowing the METADATA_MODEL to be passed into the read routine, so that it is more consistently treated.
This still has the issue that reading the DEPEND_0 variable directly will not recognize the ISTP metadata. The logic should probably be that if any variable contains DEPEND_0, then the entire file is ISTP_METADATA.
I sent some more infor related to this in email. But on multi-d data the axes are not always set to the right value depending on the axis.
The logic now is that if any variable contains DEPEND_0, then the ISTP logic is used.