The critical part of determining whether to issue a row of data seems to be whether there are any accumulated clicks or impressions. If either is non-zero, you get a row. If both are zero, you don't.

So your results may be explained by having some keywords that had different destination (or display) URLs and having received traffic while each different URL was offered. That would mean that you got more rows of data in a Keyword+URL report than a Keyword only report.

I suspect that if you cross check by accumulating the total clicks and total impressions in each report, they'll be the same number. If they are different... that's interesting!