Zogby one-day PA samples very odd

Zogby has been ok in some primaries but not in others.

But his numbers are jumping around like crazy the last few days.

The Saturday one-day sample gave Obama a lead:

Obama. . . had a good today at 46% to Clinton's 44%

But in Zogby's Sunday one-day sample, Clinton led

53 to 38

Now, doesn't that seem impossible for Clinton to go from 44 to 53 from one day to the next and Obama from 46 to 38, again from one day to the next?  If the number of undecideds was real different, it might make sense.

But since the number of undecideds is virtually identical (10% on Saturday and 9% on Sunday), it's seems to me that there's either random sampling error or perhaps a day of the week effect.

What do you think?