create_float_source code generate many nan values in matrix files

I have found that the python code to create_source_data is generating many nan values in SAL and TEMP data of WMO matrix .mat. While in netcdf files or in matlab equivalent code to create_source_data are showing much more available numeric data.

Example float with this issue: 6901967; 6901969

The reduced amount of data will lead to poorer owc estimates.

I am using the argopy v0.1.11 version.

Start by inspecting the Python create_source_data code and comparing its SAL and TEMP output for floats 6901967 and 6901969 with the corresponding netCDF files and MATLAB implementation. Confirm where valid values become NaN and verify that the generated WMO .mat files retain comparable numeric data without reducing OWC inputs.

Written by the indexing model from the issue text.