Astronomically calibrated cyclostratigraphy relies on correct matching of observed sedimentary cycles to predicted astronomical drivers such as eccentricity, obliquity and climate-precession. However the periods of these astronomical cycles, in the past, are not perfectly known because: (i) they drift through time; (ii) they overlap; (iii) they are affected by the poorly constrained recession history of the Moon. This paper estimates the resulting uncertainties in ancient Milankovitch cycle periods and shows that they lead to: (i) problems with using Milankovitch cycles for accurate measurement of durations (potential errors are around 25% by the start of the Phanerozoic); (ii) problems with correctly identifying the Milankovitch cycles responsible for observed period ratios (e.g. the ratio for long eccentricity/short-eccentricity overlaps, within error, with the ratio for short eccentricity/precession); (iii) problems with verifying that observed cycles are Milankovitch driven at all (the probability of a random period-ratio matching a predicted Milankovitch ratio, within error, is 20-70% in the Phanerozoic). Milankovitch-derived ages and durations should therefore be treated with caution unless supported by additional information such as radiometric constraints.