Whole Number ID Problem (19 Digits)

My data source is using 19 digit integers for unique IDs. According to specification, this is fine with Power BI (19 digits being the maximum width).

When I merge one query with another using this ID for the join, I get multiple rows of data where there should be just one. If I add a filter to the first query so that it contains just one client, I see the following:

• I start with 68 rows of data for the single client, all with the same ProjectUID.
• I group by ProjectUID and a couple of other common fields and aggregate some values, which gives me one row
• Then I merge with a second query and when I expand the new fields, I now have four rows - one is the expected client, the other three are totally unrelated clients
• If I then change the ProjectUID type to text, I get the single row that I was expecting.

I am trying to avoid setting these numeric IDs to text as I believe this adversely affects processing speed and I am already reaching a point where Power BI is to slow to be useable.

I have found that if I take the last 15 digits (which in this case are still unique), then everything works as expected. This suggests that the stated 19 digit width of integer type is incorrect (or that I am misunderstanding something). The number of excess rows reduces by one for each digit I remove from the beginning of the ID. I am uncertain whether the 15 digit width is working because it works or if it is just because I only had three excess rows in that filter to begin with.

Anyone else had this issue?

Just a thought - does the 19 digit width mean 18 digits plus the +/- sign?