Excellent, it all fits.
Discharged in 1828, married the following year (when he marked)
Must be him, because of the death date. Age at decease in the later pensions record matches the age given in the GRO index. Unless it's the wrong death (seems unlikely!)
So some progress at last.