Oops.  I was off by a couple orders of magnitude on the error.  Try:

		case 1200:
			/* A factor of 1.1011011 gives about .25% error. */
			if (time_adj < 0) {
				time_adj -= (-time_adj >> 1);
				time_adj -= (-time_adj >> 3) + (-time_adj >> 6);
			} else {
				time_adj += (time_adj >> 1);
				time_adj += (time_adj >> 3) + (time_adj >> 6);
			}
			break;

B-)