I can think of two good reasons. First, unlike most European countries, US has a greater population living in poverty, and its populations is also more mobile. The poor and mobile people are less likely to be included in surveys, such as that conducted by the BLS, but they also more likely to be unemployed. This is probably the same part of the population that does not appear on the census. Since GOP so desperately fights to prevent counting those people - that tells me their numbers must be nontrivial ("do not believe anything unless it is officially denied").
So the net effect is that the US unemployement stats probably undercount the poor and unemployed more than most European stats.
The second reason pertains to the numerator of the unemployment rate. BLS uses headcount rather than full-time equivalent (FTE) to account for employment. That is, a person who is employed only 10 hours a week counts the same as the one who is employed 40 hours a week. This, BTW, is a well-known weakness of the employment stats. Thus, significant numbers of the under-employed can skew the employment rates upward, as compared to the FTE approach.
I suspect that underemployment (part-tiem, temporary arrangements) are more prevalent in the US than in Europe, and that again skews the US stats upward as compared to Europe.
I do not know if those two biases can account for the 4-percentage points difference between the US and the Canadian/European rates, but it is likely that they are a nontrivial part of that difference.
Regards,
Wojtek Sokolowski