Your calculation of proportion of variance seems to be correct. The following example highlights that:
% Example from pcacov documentation page
load hald
covx = cov(ingredients);
[COEFF,latent,explained] = pcacov(covx);
cumsum(latent/sum(latent))
ans =
0.8660
0.9789
0.9996
1.0000
pareto(latent)
You will observe that the pareto chart shows only 95% of the cumulative distribution and therefore, 2 columns will be displayed.
Note also that "pcacov" performs principal components analysis on the covariance matrix of the input so calling "cov" inside "pcacov" does not seem necessary.
