The latest in our series of Morning Consult poll results is a bit later than normal this weekend, but I’m not taking the rap on this one. They normally have them done Saturday night or very early Sunday. Today was different for some reason. In any event, let us soldier on.

For the last two weeks, among likely voters, MC had Clinton with a head to head lead of 42-40 and 44-43, both times within the margin of error but showing a lean toward Clinton if nothing else. Having bounced back from a significant deficit, MC’s sample seemed to show Trump stalled. This week there’s an even more curious shift, particularly in light of the rest of the polls we’ve been covering. The results for the September 15 and 16 survey are in and – at least in the four way race – not much has changed for the third week in a row.

Hillary Clinton remains the top choice for voters in the presidential race, rebounding from her brief health scare at a 9/11 ceremony a week ago. She is still leading Donald Trump by 2 percentage points among likely voters.

In a new Morning Consult survey conducted Sept. 15 and Sept. 16, Clinton leads the Republican nominee, 42 percent to 40 percent — within the margin of error — among likely voters, while 8 percent opted for Libertarian nominee Gary Johnson and 3 percent chose Green Party candidate Jill Stein.

Among registered voters, the Democratic nominee leads Trump by 1 point, 39 percent to 38 percent. And in a head-to-head matchup, Clinton’s lead over Trump swells to 4 points among both likely and registered voters. Last week, she only led Trump by 1 point among likely voters and 2 points among registered voters.

That’s an odd result to be sure, given the rest of the national polls which show Trump either holding even or continuing to gain on Clinton nationally and taking the lead in several swing states. MC’s head to head match among likely voters, conversely, expands her lead to 46-42. (See likely voter crosstabs.) I can’t find a single survey, even in the most liberal leaning ones, showing anything remotely close to this or even trending in that direction.

Given the results, we’ll definitely want to head back to the crosstabs. Last week’s demographic breakdown came out as I described at the time.

Once again we may have to delve into the crosstabs for answers. The sample for their registered voter results this week comes up with a split of 38 Democrat, 32 Independent and 30 Republican. Last week it was 35/32/32. For the likely voter model things are even more strange. This week the D/I/R ratio jumped to 40/27/32. That’s a net five percent jump for the Democrats and a matching five percent drop for independents.

In this week’s crosstabs things still look rather odd. The likely voter D/I/R split without leaners is 39/29/32. Even with the Indies suppressed well below the national average and a D+7 break, Trump is within one in the four way race. Taking Gary Johnson and Jill Stein out of the race shifts the head to head match to a four point lead for Clinton? So in other words we had a bunch of voters who like Jill Stein but then figured that Donald Trump would be their next best match?

We’ll have to match this up against the Monday through Thursday results in other surveys and figure out if it’s still worth tracking this one every week, but there you have it.