Swapping the DirectQuery limits
The wrong options for DirectQuery limits offer the wrong number, and some are real figures from elsewhere. The 32,764 long-text character limit is offered as a row cap, beside 100,000 and ten million. Any query or intermediate operation returning more than 1,000,000 rows fails. The per-query timeout is 4 minutes, while 30 seconds is where usability begins to degrade and 15 minutes is the minimum tile refresh interval. The default per-file connection cap is 10.
See it in a sample questionGiving Direct Lake on OneLake the behaviour of Direct Lake on SQL
The wrong options credit Direct Lake on OneLake with fallback to DirectQuery, or place the restrictions on the other variant. Direct Lake on SQL is the variant that falls back to DirectQuery when a guardrail is reached, provided fallback stays enabled. On OneLake, a reached guardrail makes the model unqueryable until the Delta tables are optimised. The documented OneLake restrictions are no personal workspaces, no gateways and no cross-region models, not calculated columns or auto date/time.
See it in a sample questionMisplacing which Power Query features are Online-only
The wrong options misplace which Power Query features depend on the edition. Fuzzy grouping in the Group by dialog is only available in Power Query Online, so an analyst on Power BI Desktop cannot find the checkbox, and switching between Basic and Advanced mode or changing the column type does not change that. Don't aggregate, under the Advanced option of the pivot dialog, is the opposite case: it is available in both Online and Desktop, so an option calling it Online-only is wrong.
See it in a sample questionTrusting a data type that looks adequate
The wrong options lean on a type that looks adequate. Decimal number is a 64-bit floating-point type that cannot represent every number in its range exactly, and Percentage inherits the same concern, so currency belongs in Fixed decimal number, which is always precise. Detect data type interprets values using the document locale, so a United Kingdom day/month/year column in an English (United States) file still errors. Change type, Using locale, with English (United Kingdom) resolves it.
See it in a sample questionMisreading how Merge and Append match and update queries
The wrong options assume the wrong matching rule or the wrong target. A multi-column Merge pairs columns in selection order, so the first column selected on the left meets the first on the right. Identical column names are not required, and pre-sorting does not decide the match. Append matches columns by header name rather than position, and the query it is run from gains the appended step while the other table is unchanged.
See it in a sample question