I’ve got a question on the interface CovarianceMatrixProvider.
The Javadoc claims that for both initial and process noise matrices, the size must be at least 6x6 as the orbital parameters coefficients must always be present. However this seems contradictory to their use in Kalman filters, where one must provide difference instances of CovarianceMatrixProvider for the orbital + propagation parameters on one hand and the measurement models ones on the other hand. For the latter, I get an error in the Kalman builder if I try adding 6 to the matrix size. So is the documentation wrong (maybe not updated) or am I missing something?

to illustrate my point, here is how the UKF is set in some test:

final UnscentedKalmanEstimatorBuilder kalmanBuilder = new UnscentedKalmanEstimatorBuilder();
kalmanBuilder.unscentedTransformProvider(new MerweUnscentedTransform(8));
kalmanBuilder.addPropagationConfiguration(propagatorBuilder,
new ConstantProcessNoise(MatrixUtils.createRealMatrix(7, 7)));
kalmanBuilder.estimatedMeasurementsParameters(drivers, new ConstantProcessNoise(MatrixUtils.createRealIdentityMatrix(1)));