Saving data loses information?

[Jose]

Hello.

I have a particular case of a covariance matrix with a large condition 
number, whose inverse is problematic
---
debug> eig(cholinv(cov_b));
error: eig: EIG: matrix contains Inf or NaN values
debug> any(any(isnan(cov_b)||isinf(cov_b)))
ans = 0
debug> cond(cov_b)
ans =  1.0901e+06
---
I saved the matrix in a file using the binary option, and to my 
surprise, when loaded into another freshly opened octave session, I can 
execute eig(cholinv(cov_b)) with no problems.

Could it be the cause of this is that the saving and loading of the data 
is not 100% accurate? Is there any other possibility?

If so, is there any format that is guaranteed to be 100% accurate?

Br,
Jose