Solved

Optimizing spying strategy

The answer is unfortunately: It depends. The percentage describes how much of the difference between your cars attribute value and the one of the best team in that attribute you'll get as your research result, rounded up. Which means if those differences are rather equal, and especially if they are small, then it's better to split up. If you've got a larger deficit in one attribute it's better to concentrate on that. Also sometimes the rounding effect favours choosing two. 

Examples:
You're behind 4 4 4 points in your 3 remaining attributes, with selecting 3 and 16% it'd be a result of 0.64 which means 1 point each to a total of 3, with 2 and 23% it's 0.92 which also means 1 point each selected resulting in a total of 2 points, with 1 and 46% it's 1.84 and thus results also in 2 points. Selecting all 3 is best (actually you can take everything else you're behind in, too, because it doesn't matter if it's 0.24 or 0.64, you get 1 point for your main attributes either way, but with selecting more you reap some extra points in the minor stats)

But 5 5 5 would give 3 * 1 (0.8) = 3, 2 * 2 (1.15) = 4 and 3 (2.3) by selecting one. Selecting 2 wins the day

Now 10 4 and 3, selecting 3 gives 2 (1.6) plus 1 (0.64) and 1 (0.48) equals 4, 2 is 3 (2.3) and 1 (0.92) equals 4, but 1 gives still 5 points (4.6). Selecting the highest one only is best

No, exact numbers on the gap are also not available in a browser. You can only guess/measure the gap from the bars, just like in the app.

edit: 39 seconds slower than Kevin ;(